Social deduction games like Werewolf combine language, reasoning, and strategy, providing a testbed for studying natural language and social intelligence. However, most studies reduce the game to LLM-based self-play, yielding templated utterances and anecdotal cases that overlook the richness of social gameplay. Evaluation further relies on coarse metrics such as survival time or subjective scoring due to the lack of quality reference data. To address these gaps, we curate a high-quality, human-verified multimodal Werewolf dataset containing over 100 hours of video, 32.4M utterance tokens, and 15 rule variants. Based on this dataset, we propose a novel strategy-alignment evaluation that leverages the winning faction's strategies as ground truth in two stages: 1) Speech evaluation, formulated as multiple-choice-style tasks that assess whether the model can adopt appropriate stances across five dimensions of social ability; and 2) Decision evaluation, which assesses the model's voting choices and opponent-role inferences. This framework enables a fine-grained evaluation of models' linguistic and reasoning capabilities, while capturing their ability to generate strategically coherent gameplay. Our experiments show that state-of-the-art LLMs show diverse performance, with roughly half remain below 0.50, revealing clear gaps in deception and counterfactual reasoning. We hope our dataset further inspires research on language, reasoning, and strategy in multi-agent interaction.

Anemia, a condition marked by insufficient levels of red blood cells or hemoglobin, remains a widespread health issue affecting millions of individuals globally. Accurate and timely diagnosis is essential for effective management and treatment of anemia. In recent years, there has been a growing interest in the use of artificial intelligence techniques, i.e., machine learning (ML) and deep learning (DL) for the detection, classification, and diagnosis of anemia. This paper provides a systematic review of the recent advancements in this field, with a focus on various models applied to anemia detection. The review also compares these models based on several performance metrics, including accuracy, sensitivity, specificity, and precision. By analyzing these metrics, the paper evaluates the strengths and limitation of discussed models in detecting and classifying anemia, emphasizing the importance of addressing these factors to improve diagnostic accuracy.

Most defences target the pre-training stage, leaving memorisation during fine-tuning--especially for domain adaptation and instruction tuning--poorly understood. We fine-tune Pythia, Llama3, and Mistral models spanning 1.4B-70B parameters on common evaluation datasets and track verbatim memorisation throughout training. We find that memorisation increases dramatically in the first few epochs, often significantly before either validation perplexity or evaluation performance is op-timised. We use a simple but effective n-gram memorisation score which reliably precedes verbatim memorisation; using it as an early-stopping criterion mitigates memorisation with minimal performance loss. Further, we introduce an n-gram-aware loss regulariser and show that it reduces memorisation across all model families tested by up to 40% while minimising evaluation performance trade-offs when compared to an existing memorisation mitigation strategy. These results yield practical, scalable insights into memorisation dynamics during language model fine-tuning.

Style transfer combines the content of one signal with the style of another. It supports applications such as data augmentation and scenario simulation, helping machine learning models generalize in data-scarce domains. While well developed in vision and language, style transfer methods for time series data remain limited. We introduce DiffTSST, a diffusion-based framework that disentangles a time series into content and style representations via convolutional encoders and recombines them through a self-supervised attention-based diffusion process. At inference, encoders extract content and style from two distinct series, enabling conditional generation of novel samples to achieve style transfer. We demonstrate both qualitatively and quantitatively that DiffTSST achieves effective style transfer. We further validate its real-world utility by showing that data augmentation with DiffTSST improves anomaly detection in data-scarce regimes.

Contrastive audio-language pretraining yields powerful joint representations, yet a persistent audio-text modality gap limits the benefits of coupling multimodal encoders with large language models (LLMs). We present Diffusion-Link, a diffusion-based modality-bridging module that generatively maps audio embeddings into the text-embedding distribution. The module is trained at the output embedding from the frozen multimodal encoder and implemented as a lightweight network with three residual MLP blocks. To assess the effect of Diffusion-Link on multimodal encoder-LLM coupling, we evaluate on Automatic Audio Captioning (AAC); to our knowledge, this is the first application of diffusion-based modality bridging to AAC. We report two results. (1) Modality-gap analysis: on similarity and geometric criteria, Diffusion-Link reduces the modality gap the most among prior diffusion-based methods and shows a collective migration of audio embeddings toward the text distribution. (2) Downstream AAC: attaching Diffusion-Link to the same multimodal LLM baseline achieves state-of-the-art on AudioCaps in both zero-shot and fully supervised captioning without external knowledge, with relative gains up to 52.5% and 7.5%, respectively. These findings show that closing the modality gap is pivotal for effective coupling between multimodal encoders and LLMs, and diffusion-based modality bridging offers a promising direction beyond knowledge-retrieval-centric designs. Code will be released upon acceptance https://github.com/DevKiHyun/Diffusion-Link

As the demand for emotional intelligence in large language models (LLMs) grows, a key challenge lies in understanding the internal mechanisms that give rise to emotional expression and in controlling emotions in generated text. This study addresses three core questions: (1) Do LLMs contain context-agnostic mechanisms shaping emotional expression? (2) What form do these mechanisms take? (3) Can they be harnessed for universal emotion control? We first construct a controlled dataset, SEV (Scenario-Event with Valence), to elicit comparable internal states across emotions. Subsequently, we extract context-agnostic emotion directions that reveal consistent, cross-context encoding of emotion (Q1). We identify neurons and attention heads that locally implement emotional computation through analytical decomposition and causal analysis, and validate their causal roles via ablation and enhancement interventions. Next, we quantify each sublayer's causal influence on the model's final emotion representation and integrate the identified local components into coherent global emotion circuits that drive emotional expression (Q2). Directly modulating these circuits achieves 99.65% emotion-expression accuracy on the test set, surpassing prompting- and steering-based methods (Q3). To our knowledge, this is the first systematic study to uncover and validate emotion circuits in LLMs, offering new insights into interpretability and controllable emotional intelligence.

This paper investigates automated skill decomposition using Large Language Models (LLMs) and proposes a rigorous, ontology-grounded evaluation framework. Our framework standardizes the pipeline from prompting and generation to normalization and alignment with ontology nodes. To evaluate outputs, we introduce two metrics: a semantic F1-score that uses optimal embedding-based matching to assess content accuracy, and a hierarchy-aware F1-score that credits structurally correct placements to assess granularity. We conduct experiments on ROME-ESCO-DecompSkill, a curated subset of parents, comparing two prompting strategies: zero-shot and leakage-safe few-shot with exemplars. Across diverse LLMs, zero-shot offers a strong baseline, while few-shot consistently stabilizes phrasing and granularity and improves hierarchy-aware alignment. A latency analysis further shows that exemplar-guided prompts are competitive - and sometimes faster - than unguided zero-shot due to more schema-compliant completions. Together, the framework, benchmark, and metrics provide a reproducible foundation for developing ontology-faithful skill decomposition systems.

Current approaches to embodied AI tend to learn policies from expert demonstrations. However, without a mechanism to evaluate the quality of demonstrated actions, they are limited to learning from optimal behaviour, or they risk replicating errors and inefficiencies. While reinforcement learning offers one alternative, the associated exploration typically results in sacrificing data efficiency. This work explores how agents trained with imitation learning can learn robust representations from both optimal and suboptimal demonstrations when given access to constructive language feedback as a means to contextualise different modes of behaviour. We directly provide language feedback embeddings as part of the input sequence into a Transformer-based policy, and optionally complement the traditional next action prediction objective with auxiliary self-supervised learning objectives for feedback prediction. We test our approach on a range of embodied Vision-and-Language tasks in our custom BabyAI-XGen environment and show significant improvements in agents' compositional generalisation abilities and robustness, suggesting that our data-efficient method allows models to successfully convert suboptimal behaviour into learning opportunities. Overall, our results suggest that language feedback is a competitive and intuitive alternative to intermediate scalar rewards for language-specified embodied tasks.

This paper proposes an agent-based approach toward a more natural interface between humans and machines. Large language models equipped with tools and the communication standard OPC UA are utilized to control machines in natural language. Instead of touch interaction, which is currently the state-of-the-art medium for interaction in operations, the proposed approach enables operators to talk or text with machines. This allows commands such as 'Please decrease the temperature by 20 % in machine 1 and set the motor speed to 5000 rpm in machine 2.' The large language model receives the user input and selects one of three predefined tools that connect to an OPC UA server and either change or read the value of a node. Afterwards, the result of the tool execution is passed back to the language model, which then provides a final response to the user. The approach is universally designed and can therefore be applied to any machine that supports the OPC UA standard. The large language model is neither fine-tuned nor requires training data, only the relevant machine credentials and a parameter dictionary are included within the system prompt. The approach is evaluated on a Siemens S7-1500 programmable logic controller with four machine parameters in a case study of fifty synthetically generated commands on five different models. The results demonstrate high success rate, with proprietary GPT 5 models achieving accuracies between 96.0 % and 98.0 %, and open-weight models reaching up to 90.0 %. The proposed approach of this empirical study contributes to advancing natural interaction in industrial human-machine interfaces.

Knowledge graphs (KGs) are vital for knowledge-intensive tasks and have shown promise in reducing hallucinations in large language models (LLMs). However, constructing high-quality KGs remains difficult, requiring accurate information extraction and structured representations that support interpretability and downstream utility. Existing LLM-based approaches often focus narrowly on entity and relation extraction, limiting coverage to sentence-level contexts or relying on predefined schemas. We propose a hierarchical extraction framework that organizes information at multiple levels, enabling the creation of semantically rich and well-structured KGs. Using state-of-the-art LLMs, we extract and construct knowledge graphs and evaluate them comprehensively from both structural and semantic perspectives. Our results highlight the strengths and shortcomings of current LLMs in KG construction and identify key challenges for future work. To advance research in this area, we also release a curated dataset of LLM-generated KGs derived from research papers on children's mental well-being. This resource aims to foster more transparent, reliable, and impactful applications in high-stakes domains such as healthcare.