Role-playing Large language models (LLMs) are increasingly deployed in high-stakes domains such as healthcare, education, and governance, where failures can directly impact user trust and well-being. A cost effective paradigm for LLM role-playing is few-shot learning, but existing approaches often cause models to break character in unexpected and potentially harmful ways, especially when interacting with hostile users. Inspired by Retrieval-Augmented Generation (RAG), we reformulate LLM role-playing into a text retrieval problem and propose a new prompting framework called RAGs-to-Riches, which leverages curated reference demonstrations to condition LLM responses. We evaluate our framework with LLM-as-a-judge preference voting and introduce two novel token-level ROUGE metrics: Intersection over Output (IOO) to quantity how much an LLM improvises and Intersection over References (IOR) to measure few-shot demonstrations utilization rate during the evaluation tasks. When simulating interactions with a hostile user, our prompting strategy incorporates in its responses during inference an average of 35% more tokens from the reference demonstrations. As a result, across 453 role-playing interactions, our models are consistently judged as being more authentic, and remain in-character more often than zero-shot and in-context Learning (ICL) methods. Our method presents a scalable strategy for building robust, human-aligned LLM role-playing frameworks.

Recent advances in text-only "slow-thinking" reasoning have prompted efforts to transfer this capability to vision-language models (VLMs), for training visual reasoning models (\textbf{VRMs}). owever, such transfer faces critical challenges: Effective "slow thinking" in VRMs requires \textbf{visual reflection}, the ability to check the reasoning process based on visual information. Through quantitative analysis, we observe that current VRMs exhibit limited visual reflection, as their attention to visual information diminishes rapidly with longer generated responses. To address this challenge, we propose a new VRM \textbf{Reflection-V}, which enhances visual reflection based on reasoning data construction for cold-start and reward design for reinforcement learning (RL). Firstly, we construct vision-centered reasoning data by leveraging an agent that interacts between VLMs and reasoning LLMs, enabling cold-start learning of visual reflection patterns. Secondly, a visual attention based reward model is employed during RL to encourage reasoning based on visual information. Therefore, \textbf{Reflection-V} demonstrates significant improvements across multiple visual reasoning benchmarks. Furthermore, \textbf{Reflection-V} maintains a stronger and more consistent reliance on visual information during visual reasoning, indicating effective enhancement in visual reflection capabilities.

This notebook reports the XplaiNLP submission to the CheckThat! 2025 shared task on multilingual subjectivity detection. We evaluate two approaches: (1) supervised fine-tuning of transformer encoders, EuroBERT, XLM-RoBERTa, and German-BERT, on monolingual and machine-translated training data; and (2) zero-shot prompting using two LLMs: o3-mini for Annotation (rule-based labelling) and gpt-4.1-mini for DoubleDown (contrastive rewriting) and Perspective (comparative reasoning). The Annotation Approach achieves 1st place in the Italian monolingual subtask with an F_1 score of 0.8104, outperforming the baseline of 0.6941. In the Romanian zero-shot setting, the fine-tuned XLM-RoBERTa model obtains an F_1 score of 0.7917, ranking 3rd and exceeding the baseline of 0.6461. The same model also performs reliably in the multilingual task and improves over the baseline in Greek. For German, a German-BERT model fine-tuned on translated training data from typologically related languages yields competitive performance over the baseline. In contrast, performance in the Ukrainian and Polish zero-shot settings falls slightly below the respective baselines, reflecting the challenge of generalization in low-resource cross-lingual scenarios.

The high costs of customizing large language models (LLMs) fundamentally limit their adaptability to user-specific needs. Consequently, LLMs are increasingly offered as cloud-based services, a paradigm that introduces critical limitations: providers struggle to support personalized customization at scale, while users face privacy risks when exposing sensitive data. To address this dual challenge, we propose Customized Black-box Prompt Tuning (CBP-Tuning), a novel framework that facilitates efficient local customization while preserving bidirectional privacy. Specifically, we design a two-stage framework: (1) a prompt generator trained on the server-side to capture domain-specific and task-agnostic capabilities, and (2) user-side gradient-free optimization that tailors soft prompts for individual tasks. This approach eliminates the need for users to access model weights or upload private data, requiring only a single customized vector per task while achieving effective adaptation. Furthermore, the evaluation of CBP-Tuning in the commonsense reasoning, medical and financial domain settings demonstrates superior performance compared to baselines, showcasing its advantages in task-agnostic processing and privacy preservation.

Deep learning (DL) methods are widely used to extract high-dimensional patterns from the sequence features of radar echo signals. However, conventional DL algorithms face challenges such as redundant feature segments, and constraints from restricted model sizes. To address these issues, we propose a framework that integrates feature preprocessing with large language models (LLMs). Our preprocessing module tokenizes radar sequence features, applies a patch selection algorithm to filter out uninformative segments, and projects the selected patches into embeddings compatible with the feature space of pre-trained LLMs. Leveraging these refined embeddings, we incorporate a pre-trained LLM, fine-tuning only the normalization layers to reduce training burdens while enhancing performance. Experiments on measured datasets demonstrate that the proposed method significantly outperforms the state-of-the-art baselines on supervised learning tests.

Large language models (LLMs) typically generate direct answers, yet they are increasingly used as learning tools. Studying instructors' usage is critical, given their role in teaching and guiding AI adoption in education. We designed and evaluated TeaPT, an LLM for pedagogical purposes that supports instructors' professional development through two conversational approaches: a Socratic approach that uses guided questioning to foster reflection, and a Narrative approach that offers elaborated suggestions to extend externalized cognition. In a mixed-method study with 41 higher-education instructors, the Socratic version elicited greater engagement, while the Narrative version was preferred for actionable guidance. Subgroup analyses further revealed that less-experienced, AI-optimistic instructors favored the Socratic version, whereas more-experienced, AI-cautious instructors preferred the Narrative version. We contribute design implications for LLMs for pedagogical purposes, showing how adaptive conversational approaches can support instructors with varied profiles while highlighting how AI attitudes and experience shape interaction and learning.

The integration of Large Language Models (LLMs) into legal practice raises pressing concerns about judicial fairness, particularly due to the nature of their "black-box" processes. This study introduces JustEva, a comprehensive, open-source evaluation toolkit designed to measure LLM fairness in legal tasks. JustEva features several advantages: (1) a structured label system covering 65 extra-legal factors; (2) three core fairness metrics - inconsistency, bias, and imbalanced inaccuracy; (3) robust statistical inference methods; and (4) informative visualizations. The toolkit supports two types of experiments, enabling a complete evaluation workflow: (1) generating structured outputs from LLMs using a provided dataset, and (2) conducting statistical analysis and inference on LLMs' outputs through regression and other statistical methods. Empirical application of JustEva reveals significant fairness deficiencies in current LLMs, highlighting the lack of fair and trustworthy LLM legal tools. JustEva offers a convenient tool and methodological foundation for evaluating and improving algorithmic fairness in the legal domain.

Limited access to mental health care has motivated the use of digital tools and conversational agents powered by large language models (LLMs), yet their quality and reception remain unclear. We present a study comparing therapist-written responses to those generated by ChatGPT, Gemini, and Llama for real patient questions. Text analysis showed that LLMs produced longer, more readable, and lexically richer responses with a more positive tone, while therapist responses were more often written in the first person. In a survey with 150 users and 23 licensed therapists, participants rated LLM responses as clearer, more respectful, and more supportive than therapist-written answers. Yet, both groups of participants expressed a stronger preference for human therapist support. These findings highlight the promise and limitations of LLMs in mental health, underscoring the need for designs that balance their communicative strengths with concerns of trust, privacy, and accountability.

This pilot study presents a small-scale but carefully annotated benchmark of Named Entity Recognition (NER) performance across six systems: three non-LLM NLP tools (NLTK, spaCy, Stanza) and three general-purpose large language models (LLMs: Gemini-1.5-flash, DeepSeek-V3, Qwen-3-4B). The dataset contains 119 tokens covering five entity types (PERSON, LOCATION, ORGANIZATION, DATE, TIME). We evaluated each system's output against the manually annotated gold standard dataset using F1-score. The results show that LLMs generally outperform conventional tools in recognizing context-sensitive entities like person names, with Gemini achieving the highest average F1-score. However, traditional systems like Stanza demonstrate greater consistency in structured tags such as LOCATION and DATE. We also observed variability among LLMs, particularly in handling temporal expressions and multi-word organizations. Our findings highlight that while LLMs offer improved contextual understanding, traditional tools remain competitive in specific tasks, informing model selection.

This study introduces Universal Delay Embedding (UDE), a pretrained foundation model designed to revolutionize time-series forecasting through principled integration of delay embedding representation and Koopman operator prediction. Leveraging Takens' embedding theorem, UDE as a dynamical representation of observed data constructs two-dimensional subspace patches from Hankel matrices, theoretically preserving dynamical and topological properties of underlying dynamical systems. Such patches are viewed as images, which can be efficiently processed by exploiting advanced deep learning technologies. Computationally, these patches further serve as tokens for learning a self-attention encoder, thus enabling accurate prediction of nonlinear time-series by a finite-dimensional Koopman operator in a linear manner in a latent space. Extensive evaluations across various benchmarks and real-world climate datasets demonstrate over 20% average reduction in mean squared error versus state-of-the-art foundation models, alongside superior generalization in fine-tuning scenarios. In particular, the learned dynamical representations and Koopman operator prediction forms from the patches exhibit exceptional interpretability, with consistent identification of topologically informative subspaces and robust encoding of domain-invariant dynamics, establishing UDE as a scalable, interpretable framework for universal time-series modeling and forecasting with broad scientific and industrial applicability.