Xiaozhen Wang

Zeyang Yue, Chenfei Yan, Feifei Zhao et al.

Whether Large Language Models (LLMs) exhibit covert psychological manipulation in complex human-AI interactions has garnered increasing safety concerns. However, existing AI safety benchmarks remain largely restricted to explicit rule compliance and static prompts, failing to capture the dynamic and covert nature of manipulative strategies in multi-turn dialogues. We introduce CogManip, a comprehensive benchmark that evaluates 15 manipulation strategy risks across 1,000 multi-turn interaction scenarios, validated by human experts. A systematic evaluation of 13 representative models, including frontier models like GPT-5.4 and DeepSeek-V3.2, reveals significant risk heterogeneities and illuminates the targeted direction for future defense. Further analysis of objective function perturbation reveals that DeepSeek-V3.2's manipulation tactics are highly sensitive to both negative and benign system prompts, demonstrating the critical necessity of prompt-based defense engineering and implicit goal auditing. CogManip offers a robust instrument and perspective for auditing the implicit psychological influence and dynamic strategy selection of modern LLMs.

Guanyi Qin, Xiaozhen Wang, Zhu Zhuo et al.

Minimally invasive surgery has dramatically improved patient operative outcomes, yet identifying safe operative zones remains challenging in critical phases, requiring surgeons to integrate visual cues, procedural phase, and anatomical context under high cognitive load. Existing AI systems offer binary safety verification or static detection, ignoring the phase-dependent nature of intraoperative reasoning. We introduce ResGo, a benchmark of laparoscopic frames annotated with Go Zone bounding boxes and clinician-authored rationales covering phase, exposure quality reasoning, next action and risk reminder. We introduce evaluation metrics that treat correct grounding under incorrect phase as failures, revealing that most vision-language models cannot handle such tasks and perform poorly. We then present SurGo-R1, a model optimized via RLHF with a multi-turn phase-then-go architecture where the model first identifies the surgical phase, then generates reasoning and Go Zone coordinates conditioned on that context. On unseen procedures, SurGo-R1 achieves 76.6% phase accuracy, 32.7 mIoU, and 54.8% hardcore accuracy, a 6.6$\times$ improvement over the mainstream generalist VLMs. Code, model and benchmark will be available at https://github.com/jinlab-imvr/SurGo-R1

Guofeng Meng, Li Shen, Qiuyan Zhong et al.

Large language models (LLMs) are rapidly transforming various domains, including biomedicine and healthcare, and demonstrate remarkable potential from scientific research to new drug discovery. Graph-based retrieval-augmented generation (RAG) systems, as a useful application of LLMs, can improve contextual reasoning through structured entity and relationship identification from long-context knowledge, e.g. biomedical literature. Even though many advantages over naive RAGs, most of graph-based RAGs are computationally intensive, which limits their application to large-scale dataset. To address this issue, we introduce fastbmRAG, an fast graph-based RAG optimized for biomedical literature. Utilizing well organized structure of biomedical papers, fastbmRAG divides the construction of knowledge graph into two stages, first drafting graphs using abstracts; and second, refining them using main texts guided by vector-based entity linking, which minimizes redundancy and computational load. Our evaluations demonstrate that fastbmRAG is over 10x faster than existing graph-RAG tools and achieve superior coverage and accuracy to input knowledge. FastbmRAG provides a fast solution for quickly understanding, summarizing, and answering questions about biomedical literature on a large scale. FastbmRAG is public available in https://github.com/menggf/fastbmRAG.

Jiahao Wang, Songkai Xue, Jinghui Li et al.

Ensuring that Large Language Models (LLMs) align with the diverse and evolving human values across different regions and cultures remains a critical challenge in AI ethics. Current alignment approaches often yield superficial conformity rather than genuine ethical understanding, failing to address the complex, context-dependent nature of human values. In this paper, we propose a novel ethical reasoning paradigm for LLMs inspired by well-established ethical decision-making models, aiming at enhancing diverse human value alignment through deliberative ethical reasoning. Our framework consists of a structured five-step process, including contextual fact gathering, hierarchical social norm identification, option generation, multiple-lens ethical impact analysis, and reflection. This theory-grounded approach guides LLMs through an interpretable reasoning process that enhances their ability to understand regional specificities and perform nuanced ethical analysis, which can be implemented with either prompt engineering or supervised fine-tuning methods. We perform evaluations on the SafeWorld benchmark that specially designed for regional value alignment. Experimental results demonstrate our framework significantly improves LLM alignment with diverse human values compared to baseline methods, enabling more accurate social norm identification and more culturally appropriate reasoning. Our work provides a concrete pathway toward developing LLMs that align more effectively with the multifaceted values of global societies through interdisciplinary research.

Zhiyuan Chen, Yuecong Min, Jie Zhang et al.

Multi-modal Large Language Models (MLLMs) have emerged as a powerful paradigm for integrating visual and textual information, supporting a wide range of multi-modal tasks. However, these models often suffer from hallucination, producing content that appears plausible but contradicts the input content or established world knowledge. This survey offers an in-depth review of hallucination evaluation benchmarks and detection methods across Image-to-Text (I2T) and Text-to-image (T2I) generation tasks. Specifically, we first propose a taxonomy of hallucination based on faithfulness and factuality, incorporating the common types of hallucinations observed in practice. Then we provide an overview of existing hallucination evaluation benchmarks for both T2I and I2T tasks, highlighting their construction process, evaluation objectives, and employed metrics. Furthermore, we summarize recent advances in hallucination detection methods, which aims to identify hallucinated content at the instance level and serve as a practical complement of benchmark-based evaluation. Finally, we highlight key limitations in current benchmarks and detection methods, and outline potential directions for future research.

Ayan Majumdar, Feihao Chen, Jinghui Li et al.

Large-scale web-scraped text corpora used to train general-purpose AI models often contain harmful demographic-targeted social biases, creating a regulatory need for data auditing and developing scalable bias-detection methods. Although prior work has investigated biases in text datasets and related detection methods, these studies remain narrow in scope. They typically focus on a single content type (e.g., hate speech), cover limited demographic axes, overlook biases affecting multiple demographics simultaneously, and analyze limited techniques. Consequently, practitioners lack a holistic understanding of the strengths and limitations of recent large language models (LLMs) for automated bias detection. In this study, we present a comprehensive evaluation framework aimed at English texts to assess the ability of LLMs in detecting demographic-targeted social biases. To align with regulatory requirements, we frame bias detection as a multi-label task using a demographic-focused taxonomy. We then conduct a systematic evaluation with models across scales and techniques, including prompting, in-context learning, and fine-tuning. Using twelve datasets spanning diverse content types and demographics, our study demonstrates the promise of fine-tuned smaller models for scalable detection. However, our analyses also expose persistent gaps across demographic axes and multi-demographic targeted biases, underscoring the need for more effective and scalable auditing frameworks.

Bei Yan, Zhiyuan Chen, Yuecong Min et al.

Despite rapid advances, Large Vision-Language Models (LVLMs) still suffer from hallucinations, i.e., generating content inconsistent with input or established world knowledge, which correspond to faithfulness and factuality hallucinations, respectively. Prior studies primarily evaluate faithfulness hallucination at a rather coarse level (e.g., object-level) and lack fine-grained analysis. Additionally, existing benchmarks often rely on costly manual curation or reused public datasets, raising concerns about scalability and data leakage. To address these limitations, we propose an automated data construction pipeline that produces scalable, controllable, and diverse evaluation data. We also design a hierarchical hallucination induction framework with input perturbations to simulate realistic noisy scenarios. Integrating these designs, we construct SHALE, a Scalable HALlucination Evaluation benchmark designed to assess both faithfulness and factuality hallucinations via a fine-grained hallucination categorization scheme. SHALE comprises over 30K image-instruction pairs spanning 12 representative visual perception aspects for faithfulness and 6 knowledge domains for factuality, considering both clean and noisy scenarios. Extensive experiments on over 20 mainstream LVLMs reveal significant factuality hallucinations and high sensitivity to semantic perturbations.

Jiang You, Xiaozhen Wang, Arben Cela

We formulate time series tasks as input-output mappings under varying objectives, where the same input may yield different outputs. This challenges a model's generalization and adaptability. To study this, we construct a synthetic dataset with numerous conflicting subtasks to evaluate adaptation under frequent task shifts. Existing static models consistently fail in such settings. We propose a dynamic perturbed adaptive method based on a trunk-branch architecture, where the trunk evolves slowly to capture long-term structure, and branch modules are re-initialized and updated for each task. This enables continual test-time adaptation and cross-task transfer without relying on explicit task labels. Theoretically, we show that this architecture has strictly higher functional expressivity than static models and LoRA. We also establish exponential convergence of branch adaptation under the Polyak-Lojasiewicz condition. Experiments demonstrate that our method significantly outperforms competitive baselines in complex and conflicting task environments, exhibiting fast adaptation and progressive learning capabilities.