ABSTRACT DeepSeek - R1 is a cutting - edge open - source large language model (LLM) developed by DeepSeek, showcasing advanced reasoning capabilities through a hybrid architecture that integrates m ixture of e xperts (MoE), chain of thought (CoT) reasoning, and reinforcement learning. Released under the per missive MIT license, DeepSeek - R1 offers a transparent and cost - effective alternative to proprietary models like GPT - 4o and Claude - 3 Opus; i t excels in structured problem - solving domains such as mathematics, healthcare diagnostics, code generation, and phar maceutical research. Its architecture enables efficient inference while preserving reasoning depth, making it suitable for deployment in resource - constrained settings. However, DeepSeek - R1 also exhibits increased vulnerability to bias, misinformat ion, adversarial manipulation, and safety failures - especially in multilingual and ethically sensitive contexts. Th is survey highlights the model's strengths, including interpretability, scalability, and adaptability, alongside its limitations in general language fluency and safety alignment. Future research priorities include improving bias mitigation, natural language compreh ension, domain - specific validation, and regulatory compliance. Overall, DeepSeek - R1 represents a major advance in open, scalable AI, underscoring the need for collaborative governance to ensure responsible and equitable deployment. INTRODUCTION T he rise of AI and generative models in health and technology Artificial Intelligence (AI) has undergone transformative growth in recent years, profoundly reshaping numerous fields including language processing, automation, and complex decision - making. At its core, AI refers to the simulation of human intelligence by machines, enabling them to perform tasks such as speech recognition, natural lang uage understanding, visual perception, and predictive analytics. One of the recent remarkable advancements in the Generative AI domain is the emergence of DeepSeek - R1, a large language model (LLM) developed by the Chinese company DeepSeek. In benchmarking evaluations, it has demonstrated results competitive with, and in some domains superior to, models like OpenAI's GPT - 4o and GPT - o1 [4] . This has positioned DeepSeek - R1 as a notable advancement not only in LLM capability but also in the global AI development race. DeepSeek - R1: a paradigm shift in LLM development What sets DeepSeek - R1 apart from conventional LLMs is its novel training architecture. This hybrid approach mimics certain aspects of human learning, allowing the model to refine its behavior over time and adapt to mo re complex reasoning tasks.

Knowing which latent conditions lead to a particular outcome is useful for critically examining claims made about complex event outcomes. Identifying implied conditions and examining their influence on an outcome is challenging. We handle this by combining and augmenting annotations from two existing datasets consisting of goals and states, and explore the influence of conditions through our research questions and Condition-based Reasoning tasks. We examine open and closed LLMs of varying sizes and intent-alignment on our reasoning tasks and find that conditions are useful when not all context is available. Models differ widely in their ability to generate and identify outcome-variant conditions which affects their performance on outcome validation when conditions are used to replace missing context. Larger models like GPT-4o, are more cautious in such less constrained situations.

Traditional Chinese Medicine (TCM) is a holistic medical system with millennia of accumulated clinical experience, playing a vital role in global healthcare-particularly across East Asia. However, the implicit reasoning, diverse textual forms, and lack of standardization in TCM pose major challenges for computational modeling and evaluation. Large Language Models (LLMs) have demonstrated remarkable potential in processing natural language across diverse domains, including general medicine. Yet, their systematic evaluation in the TCM domain remains underdeveloped. Existing benchmarks either focus narrowly on factual question answering or lack domain-specific tasks and clinical realism. To fill this gap, we introduce MTCMB-a Multi-Task Benchmark for Evaluating LLMs on TCM Knowledge, Reasoning, and Safety. Developed in collaboration with certified TCM experts, MTCMB comprises 12 sub-datasets spanning five major categories: knowledge QA, language understanding, diagnostic reasoning, prescription generation, and safety evaluation. The benchmark integrates real-world case records, national licensing exams, and classical texts, providing an authentic and comprehensive testbed for TCM-capable models. Preliminary results indicate that current LLMs perform well on foundational knowledge but fall short in clinical reasoning, prescription planning, and safety compliance. These findings highlight the urgent need for domain-aligned benchmarks like MTCMB to guide the development of more competent and trustworthy medical AI systems. All datasets, code, and evaluation tools are publicly available at: https://github.com/Wayyuanyuan/MTCMB.

Steering vision foundation models at inference time without retraining or access to large labeled datasets is a desirable yet challenging objective, particularly in dynamic or resource-constrained settings. In this paper, we introduce Visual Sparse Steering (VS2), a lightweight, test-time method that guides vision models using steering vectors derived from sparse features learned by top-$k$ Sparse Autoencoders without requiring contrastive data. Specifically, VS2 surpasses zero-shot CLIP by 4.12% on CIFAR-100, 1.08% on CUB-200, and 1.84% on Tiny-ImageNet. We further propose VS2++, a retrieval-augmented variant that selectively amplifies relevant sparse features using pseudo-labeled neighbors at inference time. With oracle positive/negative sets, VS2++ achieves absolute top-1 gains over CLIP zero-shot of up to 21.44% on CIFAR-100, 7.08% on CUB-200, and 20.47% on Tiny-ImageNet. Interestingly, VS2 and VS2++ raise per-class accuracy by up to 25% and 38%, respectively, showing that sparse steering benefits specific classes by disambiguating visually or taxonomically proximate categories rather than providing a uniform boost. Finally, to better align the sparse features learned through the SAE reconstruction task with those relevant for downstream performance, we propose Prototype-Aligned Sparse Steering (PASS). By incorporating a prototype-alignment loss during SAE training, using labels only during training while remaining fully test-time unsupervised, PASS consistently, though modestly, outperforms VS2, achieving a 6.12% gain over VS2 only on CIFAR-100 with ViT-B/32.

We introduce the $\underline{Ko}rean \underline{G}rammar \underline{E}valuation Bench\underline{M}ark (KoGEM)$, designed to assess the linguistic competence of LLMs and humans in Korean. KoGEM consists of 1.5k multiple-choice QA pairs covering five main categories and 16 subcategories. The zero-shot evaluation of 27 LLMs of various sizes and types reveals that while LLMs perform remarkably well on straightforward tasks requiring primarily definitional knowledge, they struggle with tasks that demand the integration of real-world experiential knowledge, such as phonological rules and pronunciation. Furthermore, our in-depth analysis suggests that incorporating such experiential knowledge could enhance the linguistic competence of LLMs. With KoGEM, we not only highlight the limitations of current LLMs in linguistic competence but also uncover hidden facets of LLMs in linguistic competence, paving the way for enhancing comprehensive language understanding. Our code and dataset are available at: https://github.com/SungHo3268/KoGEM.

Detecting ambiguity is important for language understanding, including uncertainty estimation, humour detection, and processing garden path sentences. We assess language models' sensitivity to ambiguity by introducing an adversarial ambiguity dataset that includes syntactic, lexical, and phonological ambiguities along with adversarial variations (e.g., word-order changes, synonym replacements, and random-based alterations). Our findings show that direct prompting fails to robustly identify ambiguity, while linear probes trained on model representations can decode ambiguity with high accuracy, sometimes exceeding 90\%. Our results offer insights into the prompting paradigm and how language models encode ambiguity at different layers. We release both our code and data: https://github.com/coastalcph/lm_ambiguity.

Sparse dictionary learning (and, in particular, sparse autoencoders) attempts to learn a set of human-understandable concepts that can explain variation on an abstract space. A basic limitation of this approach is that it neither exploits nor represents the semantic relationships between the learned concepts. In this paper, we introduce a modified SAE architecture that explicitly models a semantic hierarchy of concepts. Application of this architecture to the internal representations of large language models shows both that semantic hierarchy can be learned, and that doing so improves both reconstruction and interpretability. Additionally, the architecture leads to significant improvements in computational efficiency.

LoRA has emerged as one of the most promising fine-tuning techniques, especially for federated learning (FL), since it significantly reduces communication and computation costs at resource-constrained clients. However, data heterogeneity remains a significant challenge for LoRA-based FL, and the conventional aggregation strategy based on FedAvg suffers from slow convergence and suboptimal accuracy. Motivated by recent advances in model merging, particularly Task Arithmetic, we explore the idea of aggregating client LoRA parameters using scaled averaging. We first observe that a naive application of Task Arithmetic is ineffective due to the high cosine similarity between client updates, indicating significant common knowledge in the updates across clients. To address this issue, we propose decomposing client LoRA updates via Robust Principal Component Analysis (Robust-PCA) into a common low-rank component and client-specific sparse components. Our proposed algorithm FedRPCA aggregates the low-rank components through averaging, consolidating common knowledge, and applies scaled averaging to the sparse components to amplify client-specific knowledge. We evaluate our approach across a variety of vision and language tasks and demonstrate that it achieves higher final accuracy and faster convergence compared to competing baselines.

Addressing this gap is crucial for equitable AI deployment. We introduce Culturally-Grounded Chain-of-Thought (CG-CoT), a novel prompting strategy that combines dense vector retrieval of cultural context with explicit reasoning sequences. Our extensive experiments on Yoruba proverb interpretation demonstrate that CG-CoT provides significantly higher culturally-aligned accuracy and depth than traditional prompting methods, validated through both automated metrics and LLM-based evaluations. Notably, we uncover stark disparities between token-level translation metrics like BLEU and human-judged cultural relevance, suggesting a rethinking of evaluation approaches for low-resource NLP.

Grounded text generation models often produce content that deviates from their source material, requiring user verification to ensure accuracy. Existing attribution methods associate entire sentences with source documents, which can be overwhelming for users seeking to fact-check specific claims. In contrast, existing sub-sentence attribution methods may be more precise but fail to align with users' interests. In light of these limitations, we introduce Localized Attribution Queries (LAQuer), a new task that localizes selected spans of generated output to their corresponding source spans, allowing fine-grained and user-directed attribution. We compare two approaches for the LAQuer task, including prompting large language models (LLMs) and leveraging LLM internal representations. We then explore a modeling framework that extends existing attributed text generation methods to LAQuer. We evaluate this framework across two grounded text generation tasks: Multi-document Summarization (MDS) and Long-form Question Answering (LFQA). Our findings show that LAQuer methods significantly reduce the length of the attributed text. Our contributions include: (1) proposing the LAQuer task to enhance attribution usability, (2) suggesting a modeling framework and benchmarking multiple baselines, and (3) proposing a new evaluation setting to promote future research on localized attribution in content-grounded generation.