Papers

The rapid evolution of artificial intelligence in molecular science necessitates a shift from data-driven predictions to knowledge-guided reasoning. Existing molecular models are predominantly proprietary, lacking general molecular intelligence and generalizability. To address this, we propose a task-adaptive large reasoning model that integrates molecular scientific logic to emulate the thinking of molecular scientists, with capabilities for reasoning and reflection. Our approach incorporates multi-specialist modules to provide versatile molecular expertise and a chain-of-thought (CoT) framework enhanced by reinforcement learning infused with molecular knowledge, enabling structured and reflective reasoning. The model outperforms over 20 state-of-the-art multi-task large language models (LLMs) across 10 molecular tasks on 47 metrics, including property prediction, molecule generation, and reaction prediction.It achieves a 50.3% improvement over the base model while ensuring interpretability. It can bridge data-driven and knowledge-integrated approaches for intelligent molecular design.

Recently developed fused-ring organic electron-acceptors such as Y6 have strong oscillator strength, good charge-carrier transport and low bandgaps. They therefore have enormous current technical application to optoelectronic devices, such as solar cells. Due to the large number of atoms involved in representative aggregates of these materials, we need an efficient electronic structure method to model them. Standard density functional theory poorly describe charge-transfer states, and were developed for vacuum calculations of individual molecules. In this work we tune a range-separated hybrid functional for Y6. We characterise representative dimers of the solid-state and show that Y6 dimers show the extensive solvatochromic effects are due, in part, to oscillator strength borrowing. We provide an explanation for the short optimally tuned range-separation parameter, based in the Penn model for the frequency dependent dielectric of a semiconductor. We caution that standard range-separated hybrids are less accurate than global hybrids for these, and similar, materials. We show how reducing the range-separation length improves the accuracy of standard functionals, without an involved tuning process.

Recommendation systems (RS) help users find interested content and connect authors with their target audience. Most research in RS tends to focus either on predicting users’ immediate feedback (like click-through rate) accurately or improving users’ long-term engagement. However, they ignore the influence for authors and the lifelong interaction value (LIV) of user-author pairs, which is particularly crucial for improving the prosperity of social community on different platforms. Currently, reinforcement learning (RL) can optimize long-term benefits and has been widely applied in RS. In this paper, we introduce RL to Reinforce Lifelong Interaction Value of User-Author pairs (RLIV-UA) based on each interaction of UA pairs. To address the long intervals between UA interactions and the large scale of the UA space, we propose a novel Sparse Cross-Request Interaction Markov Decision Process (SCRI-MDP) and introduce an Adjacent State Approximation (ASA) method to construct RL training samples. Additionally, we introduce Multi-Task Critic Learning (MTCL) to capture the progressive nature of UA interactions (click → follow → gift), where denser interaction signals are leveraged to compensate for the learning of sparse labels. Finally, an auxiliary supervised learning task is designed to enhance the convergence of the RLIV-UA model. In offline experiments and online A/B tests, the RLIV-UA model achieves both higher user satisfaction and higher platform profits than compared methods.

Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the 'where' and 'how' questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs.

The convergence of artificial intelligence and systems biology is giving rise to a new paradigm in biomedical research—AI-powered virtual biological systems. From single-cell simulations to organ-level models and ultimately programmable virtual humans, this digital continuum holds transformative potential for disease modeling, personalized medicine, and therapeutic discovery. In this review, we critically examine the state of the art in AI-driven simulations, including the numerical foundations, multiscale integration strategies, and the emerging class of hybrid models that bridge mechanistic and data-driven approaches. We explore the challenges of validation, uncertainty quantification, and regulatory alignment across simulation scales, with particular focus on the development of simulation accountability frameworks such as SIM-CARDs. Ethical and privacy concerns, including algorithmic bias and data sovereignty in patient-specific models, are also addressed, alongside concrete proposals for governance and federated simulation workflows. Special attention is given to the technical complexity of multiscale modeling, including the integration of mechanistic solvers with neural architectures and the computational resources required for real-time, clinically actionable simulations. We conclude with a translational roadmap for virtual biology that projects validated virtual cells for drug screening by 2030, multi-organ simulations by 2040, and the emergence of programmable virtual humans by 2055. By unifying high-fidelity numerical models with explainable AI, and aligning simulation design with ethical, regulatory, and clinical needs, the field of digital biology is positioned to unlock scalable and trustworthy biomedical innovation.

大语言模型在长程交互中面临记忆过载与用户失控的双重困境：无差别的海量存储导致认知负荷攀升，黑箱式的遗忘机制引发隐私信任危机。本研究提出一种兼具认知自适应与用户可干预的 AI 记忆管理理论框架（CAUM）。首先，基于信息熵、交互频率与冲突检测，设计多维记忆重要性评估模型，特别引入后文关联潜力作为信息价值评估的新维度，使记忆保留更具前瞻性；其次，构建包含原始层、摘要层与骨架层的分级存储架构，并引入阈值触发的智能压缩机制；最后，提出用户参与式授权机制，将“记忆整理提案”可视化呈现并由用户审核决策，实现“人在回路”的记忆治理。该框架为缓解 LLM 记忆过载问题提供了系统的概念方案，将信息生命周期理论拓展至 AI 记忆管理领域，强调用户中心的信息处置权，为人工智能时代的信息生命周期管理提供了新的理论视角，也为构建用户可控的智能记忆系统奠定了概念基础。