Goal-oriented conversational systems require making sequential decisions under uncertainty about the user's intent, where the algorithm must balance information acquisition and target commitment over multiple turns. Existing approaches address this challenge from different perspectives: structured methods enable multi-step planning but rely on predefined schemas, while LLM-based approaches support flexible interactions but lack long-horizon decision making, resulting in poor coordination between information acquisition and target commitment. To address this limitation, we formulate goal-oriented conversation as an uncertainty-aware sequential decision problem, where uncertainty serves as a guiding signal for multi-turn decision making. We propose a Conversation Uncertainty-aware Planning framework (CUP) that integrates language models with structured planning: a language model proposes feasible actions, and a planner evaluates their long-term impact on uncertainty reduction. Experiments on multiple conversational benchmarks show that CUP consistently improves success rates while requiring fewer interaction turns. Further analysis demonstrates that uncertainty-aware planning contributes to more efficient information acquisition and earlier confident commitment.

Dual-recognition elements can strengthen target binding event, avoid non-specific adsorption, and improve analytical accuracy. Surface-imprinted self-assembled monolayer (SAM) formed on the substrate surface by the co-assembly of template molecules and organic monomers is considered as a prospective alternative to the conventional molecularly imprinted polymers. Herein, we propose a strategy for protein recognition through the formation of aptamer-enabled antifouling peptide-imprinted SAM. The aptamer-protein conjugates were anchored on the gold surface, and then antifouling zwitterionic peptides were assembled around the aptamer-protein conjugates to form imprinted SAM. Removing the bound proteins by an acidic solution allowed for the formation of biocompatible cavities for target rebinding. The antifouling peptides could eliminate the non-specific adsorption and strengthen the target binding event through the formation of imprinted cavities. The dual-recognition system was used to directly detect carcinoembryonic antigen (CEA) at the concentration down to 0.1 ng/mL by electrochemical impedance spectroscopy. Furthermore, homodimeric glucose oxidase (GOx) was in-situ assembled on the electrode surface to form protein networks by using homotetramer concanavalin A (ConA) as both the recognition element and the crosslinker, thereby achieving enzymatic signal amplification. The sensitivity was improved by 100-fold through the signal amplification of ConA-GOx assemblies. The proposed strategy opens up a universal route for the design of novel biosensors for the dual-recognition and accurate detection of biomarkers, providing valuable insights into the fabrication of imprinting materials and the development of innovative biosensing platforms.

While biological and pharmaceutical knowledge networks have significantly propelled drug repurposing efforts, reliance solely on these networks is insufficient for accurately addressing genetic and phenotypic variance. This limitation highlights the need for an integrative approach that leverages context-specific data to enhance the precision of drug repurposing. We introduce a network-based integrative drug scoring approach that synergistically incorporates data-driven and knowledge-driven networks without requiring their direct integration. We developed a synergistic label propagation algorithm that facilitates information transfer from data-driven to knowledge-driven networks. To enable context-specific drug repurposing, we constructed a data-driven disease-disease association network utilizing European-specific genetic information and a knowledge-driven drug-target protein association network. In a proof-of-concept study, drug scoring was applied to identify candidate drugs for rheumatoid arthritis, asthma, and multiple sclerosis. Compared with a representative direct-integration benchmark, the proposed method achieved an average AUC of 0.701, corresponding to a 9.71

Alzheimer's disease (AD) is a progressive neurodegenerative disease with high inter-patient variance in rate of cognitive decline. AD progression prediction aims to forecast patient cognitive decline and benefits from incorporating multiple neuroimaging modalities. However, existing multimodal models fail to make accurate predictions when many modalities are missing during inference, as is often the case in clinical settings. To increase multimodal model flexibility under high modality missingness, we introduce PerM-MoE, a novel sparse mixture-of-experts method that uses independent routers for each modality in place of the conventional, single router. Using T1-weighted MRI, FLAIR, amyloid beta PET, and tau PET neuroimaging data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), we evaluate PerM-MoE, state-of-the-art Flex-MoE, and unimodal neuroimaging models on predicting two-year change in Clinical Dementia Rating-Sum of Boxes (CDR-SB) scores under varying levels of modality missingness. PerM-MoE outperforms the state of the art in most variations of modality missingness and demonstrates more effective utility of experts than Flex-MoE.

Task vectors offer a compelling mechanism for accelerating inference in in-context learning (ICL) by distilling task-specific information into a single, reusable representation. Despite their empirical success, the underlying principles governing their emergence and functionality remain unclear. This work proposes the Linear Combination Conjecture, positing that task vectors act as single in-context demonstrations formed through linear combinations of the original ones. We provide both theoretical and empirical support for this conjecture. First, we show that task vectors naturally emerge in linear transformers trained on triplet-formatted prompts through loss landscape analysis. Next, we predict the failure of task vectors on representing high-rank mappings and confirm this on practical LLMs. Our findings are further validated through saliency analyses and parameter visualization, suggesting an enhancement of task vectors by injecting multiple ones into few-shot prompts. Together, our results advance the understanding of task vectors and shed light on the mechanisms underlying ICL in transformer-based models.

Large Language Models (LLMs) excel at problem solving by generating chain of thoughts in natural language, but such verbal thinking is computationally costly and prone to overthinking. Recent work instead proposes a latent thinking architecture Huginn-3.5B, which represents intermediate reasoning steps as sequence of latent representations. However, latent thoughts lack interpretability and are difficult to supervise, raising concerns about the correctness and reliability of its latent thinking processes. In this paper, we provide a systematic study of how Huginn-3.5B thinks in the latent space and how external supervision signals can improve its latent thinking processes. We show that latent thoughts leading to correct versus incorrect answers exhibit highly distinguishable patterns, and that a latent classifier can reliably predict answer correctness directly from latent thoughts. Leveraging these insights, we propose Latent Thinking Optimization (LTO), a probabilistic algorithm that employs the latent classifier as a Latent Reward Model (LRM) to optimize the latent thinking processes. Extensive experiments across diverse reasoning tasks demonstrate that LRM is highly effective in detecting incorrect latent thinking patterns, and LTO can significantly improve the latent thinking processes. Furthermore, we show that LRM can generalize across diverse domains, and LTO can be seamlessly applied to general LLMs to improve their thinking processes. In contrast to verbal thinking, our method demonstrates that reward modeling and scaling test-time thinking with supervision can be performed directly in the latent space, highlighting its potential as a general, efficient, and domain-agnostic approach to improving the thinking processes of LLMs.

Drug discovery and development is time-consuming and resource-intensive, motivating computational approaches such as diffusion models for de novo drug design. Many such models follow the structure-based drug design (SBDD) paradigm, generating molecules to fit a target binding pocket. However, existing diffusion-based SBDD methods typically couple pocket and ligand representation learning, model interactions only at the atom level, and prioritize binding affinity over other developability properties. Here, we introduce conDitar-dev, a conditional diffusion-based SBDD framework for generating ligands with strong binding affinities and favorable ADMET properties. It consists of three modules: msPRL, a pretrained multi-scale pocket representation learning module; conDitar, a pocket-conditioned diffusion model guided by msPRL representations; and paOPT, a generation-time method for optimizing ligand developability. On a newly curated benchmark of human disease targets, conDitar outperforms state-of-the-art SBDD baselines, achieving an average binding score of -8.85 kcal/mol. Across five ADMET properties, conDitar-dev improves performance by up to 73

Alzheimer’s disease (AD) has become a neurodegenerative disease with an increasing incidence rate and a large economic and social burden worldwide. Amyloid-beta oligomer (AβO) has been confirmed as a key neurotoxic species and a core diagnostic biomarker in AD. Traditional methods for AβO detection have drawbacks, such as cumbersome operation, high cost, and dependence on sophisticated instruments, hindering their transformation into fast and real-time detection techniques. (Photo)electrochemical biosensors have attracted much attention due to their inherent advantages, such as high sensitivity, low cost, portability, and ease of miniaturization. This review systematically summarizes the latest progress of (photo)electrochemical biosensors for AβO detection, mainly based on two sensing modes: direct detection and sandwich-type detection. We comprehensively elaborated on the sensing performances and recognition elements, such as antibodies, aptamers, peptides, and molecularly imprinted polymers. The integration of functional nanomaterials and signal amplification strategies was emphasized to improve the sensitivity, selectivity, and stability of biosensors. In addition, we discussed the existing challenges and looked forward to the future development direction for the early diagnosis of AD. This article aims to provide a systematic reference for the rational design and practical application of advanced biosensors in biomarker detection and AD-related precision medicine.

Differentiating between cardiac and pulmonary diseases in emergency settings presents a significant challenge due to overlapping symptoms like dyspnea and chest pain, where misdiagnosis can lead to inappropriate interventions and increased morbidity. While electrocardiograms (ECGs) and chest X-rays (CXRs) provide complementary diagnostic information, existing multimodal fusion approaches fail to fully capture the complex relationships between these fundamentally different data modalities. To address these limitations, we propose DDMF-Net, a Dual-Domain Multimodal Fusion Network that explicitly unifies multi-domain features—from both frequency and spatial/temporal perspectives—and conducts cross-modality fusion of ECG, CXR signals and clinical parameters in hyperbolic space, thereby enhancing the modeling of complex cardiopulmonary pathophysiology. Our framework contains three innovations: (1) a frequency fusion module that captures complementary spectral patterns across modalities, (2) an inter-domain fusion module that dynamically balances domain-specific features, and (3) a hyperbolic cross-attention module with soft-entailment loss that effectively models hierarchical relationships between low-level imaging/signal data and high-level clinical parameters. Evaluated on four MIMIC datasets, DDMF-Net achieves state-of-the-art performance with over 2.9% improvement in micro-AUC, enabling more accurate differentiation of cardiac and pulmonary conditions in time-sensitive emergency settings. Code is publicly available at https://github.com/kknankk/DDMF_Net.

Alzheimer’s disease (AD) is the most common neurodegenerative disorder worldwide. Early diagnosis of AD is crucial for delaying disease progression and improving patients’ quality of life. Blood biomarkers, particularly amyloid-beta (Aβ) and Tau proteins along with their phosphorylated isoforms, show advantages such as convenient sampling, minimal invasiveness, and excellent repeatability. However, the extremely low concentrations of AD biomarkers in blood impose stringent requirements on the sensitivity, specificity, and anti-interference capability of detection methods. Optical methods provide promising analytical platforms to address these challenges in view of their intrinsic merits of high sensitivity and selectivity; rapid response; and potential for miniaturization. This review systematically summarizes the latest advances in optical methods for the detection of the two core AD blood biomarkers (Aβ and Tau), covering techniques such as colorimetry, fluorescence, chemiluminescence, surface plasmon resonance (SPR), and surface-enhanced Raman scattering (SERS). The sensing principles, design strategies, and analytical performances of these methods are discussed, with special emphasis on different signal amplification strategies. In addition, several challenges and future prospects are provided with a primary focus on single-molecule detection, insufficient sensitivity and stability, lack of validation with large clinical cohorts, and absence of standardization. This review aims to provide researchers with guidance for the rational development of high-performance optical methods to achieve early diagnosis of AD.

Multi-objective retrosynthesis planning is a critical chemistry task requiring dynamic balancing of quality, safety, and cost objectives. Language model-based multi-agent systems (MAS) offer a promising approach for this task: leveraging interactions of specialized agents to incorporate multiple objectives into retrosynthesis planning. We present MMORF, a framework for constructing MAS for multi-objective retrosynthesis planning. MMORF features modular agentic components, which can be flexibly combined and configured into different systems, enabling principled evaluation and comparison of different system designs. Using MMORF, we construct two representative MAS: MASIL and RFAS. On a newly curated benchmark consisting of 218 multi-objective retrosynthesis planning tasks, MASIL achieves strong safety and cost metrics on soft-constraint tasks, frequently Pareto-dominating baseline routes, while RFAS achieves a 48.6

Reaction feasibility prediction, as a fundamental problem in computational chemistry, has benefited from diverse tools enabled by recent advances in artificial intelligence, particularly large language models. However, the performance of individual tools varies substantially across reactions, making it difficult for any single tool to consistently perform well across all cases. This raises a critical challenge: how to effectively leverage multiple tools to obtain more accurate feasibility predictions. To address this, we propose ARMOR, an agentic framework that explicitly models tool-specific utilities, adaptively prioritizes tools, and further resolves the potential tool conflicts to produce the final prediction for each reaction. Unlike existing approaches that rely on simple aggregation or heuristic assignment over various tools, ARMOR organizes tools into a hierarchy that prioritizes top-performing tools and defers others when needed, characterizes their strengths through tool-specific patterns, and resolves conflicts via memoryaugmented reasoning. Extensive experiments on a public dataset demonstrate that ARMOR consistently outperforms strong baselines, including single-tool methods as well as various tool aggregation and tool selection approaches. Further analysis shows that the improvements are particularly significant on reactions with conflicting tool predictions, highlighting the effectiveness of ARMOR in leveraging the complementary strengths of multiple tools. The code is available via https://anonymous.4open.science/r/ARMOR-E13F.

BackgroundFree-text notes in disease intervention specialist (DIS) records may contain relevant information for sexual transmitted infection control. In their current form, the notes are not analyzable without manual reading, which is labor-intensive and prone to error.MethodsWe used natural language processing methods to analyze 2019 Ohio DIS syphilis records with nonmissing notes (n = 1987). We identified 21 topics relevant for transmission and case investigations. We manually coded these records to create "gold standard" labels for each topic (0 = topic not present, 1 = topic present), then trained machine learning models to identify the topics in the text. For models to analyze text data, the text must be converted to numbers. We explored 2 approaches to numerically represent words: (1) term frequency, inverse document frequency, which measures importance of words based on how many times they appear in a record and in the dataset as a whole, and (2) GloVe embeddings, which are numerical vectors that were developed by researchers for each word in the English language to encode its semantic meaning. We explored 3 types of statistical models (naive Bayes, support vector machine, and logistic regression) using term frequency, inverse document frequency, and 1 type of neural network model (long short-term memory [LSTM] model) using GloVe. All models were used for binary prediction (i.e., topic not present, topic present).ResultsFor most topics, the LSTM model performed the best overall in identifying topics, and the support vector machine model performed the best among the statistical models. For example, the LSTM model predicted the topic "substance use" with high accuracy (97%), sensitivity (92%), and specificity (98%). No model performed well for uncommon topics (e.g., "alcohol use" or "delays in care").ConclusionsMachine learning models performed well in identifying some topics in 2019 Ohio syphilis records. This analysis is a first step in applying natural language processing methods to making DIS notes more accessible for analysis.

Amyloid-β (Aβ) aggregates are considered as the important factors of Alzheimer’s disease (AD). Multifunctional materials have shown significant effects in the diagnosis and treatment of AD by modulating the aggregation of Aβ and production of reactive oxygen species (ROS). Compared to traditional surgical treatment and radiotherapy, phototherapy has the advantages, including short response time, significant efficacy, and minimal side effects in disease diagnosis and treatment. Recent studies have shown that local thermal energy or singlet oxygen generated by irradiating certain organic molecules or nanomaterials with specific laser wavelengths can effectively degrade Aβ aggregates and depress the generation of ROS, promoting progress in AD diagnosis and therapy. Herein, we outline the development of photothermal therapy (PTT) and photodynamic therapy (PDT) strategies for the diagnosis and therapy of AD by modulating Aβ aggregation. The materials mainly include organic photothermal agents or photosensitizers, polymer materials, metal nanoparticles, quantum dots, carbon-based nanomaterials, etc. In addition, compared to traditional fluorescent dyes, aggregation-induced emission (AIE) molecules have the advantages of good stability, low background signals, and strong resistance to photobleaching for bioimaging. Some AIE-based materials exhibit excellent photothermal and photodynamic effects, showing broad application prospects in the diagnosis and therapy of AD. We further summarize the advances in the detection of Aβ aggregates and phototherapy of AD using AIE-based materials.

Multimode immunoassays based on multiple response mechanisms have received great attention due to their capacity to effectively improve the accuracy and reliability of biosensing platforms. However, few strategies have been reported for triple-mode immunoassays due to the shortage of multifunctional sensing materials and the incompatibility of signal transduction methods in different detection modes. In this work, a fluorescent-electrochemical-colorimetric triple-mode immunoassay platform was proposed with Cu-based metal-organic frameworks (MOFs) as the signal labels. The captured Cu-MOFs were successfully decomposed under an acidic condition, leading to the release of numerous Cu2+ ions and 2-aminobenzene-1,4-dicarboxylic acid (NH2-BDC) ligands. The released NH2-BDC were determined by fluorescence titration. Meanwhile, the released Cu2+ were readily quantified by differential pulse voltammetry (DPV) and simply detected through the catalytic oxidation of chromogenic substrate 3,3',5,5'-tetramethylbenzidine (TMB). Taking alpha-fetoprotein (AFP) as a model analyte, the designed triple-mode immunoassays showed good performances with the linear range of 10-200 pg/mL, 10-200 pg/mL, and 1-100 pg/mL for the fluorescent, electrochemical, and colorimetric modes, respectively. The proposed triple-mode biosensing platforms show great potential for the applications in disease diagnosis, since they can be easily extended to other bioassays by changing the targets and recognition elements.

In real-world drug design, molecule optimization requires selectively improving multiple molecular properties up to pharmaceutically relevant levels, while maintaining others that already meet such criteria. However, existing computational approaches and instruction-tuned LLMs fail to capture such nuanced property-specific objectives, limiting their practical applicability. To address this, we introduce C-MuMOInstruct, the first instruction-tuning dataset focused on multi-property optimization with explicit, property-specific objectives. Leveraging C-MuMOInstruct, we develop \mathtt{GeLLM^4O\text{-}C} s, a series of instruction-tuned LLMs that can perform targeted property-specific optimization. Our experiments across 5 in-distribution and 5 out-of-distribution tasks show that \mathtt{GeLLM^4O\text{-}C} s consistently outperform strong baselines, achieving up to 126% higher success rate. Notably, \mathtt{GeLLM^4O\text{-}C} s exhibit impressive 0-shot generalization to novel optimization tasks and unseen instructions. This offers a step toward a foundational LLM to support realistic, diverse optimizations with property-specific objectives. C-MuMOInstruct and code are accessible through https://github.com/ninglab/GeLLMO-C.

To enhance large language models (LLMs) for chemistry problem solving, several LLM-based agents augmented with tools have been proposed, such as ChemCrow and Coscientist. However, their evaluations are narrow in scope, leaving a large gap in understanding the benefits of tools across diverse chemistry tasks. To bridge this gap, we develop ChemToolAgent, an enhanced chemistry agent over ChemCrow, and conduct a comprehensive evaluation of its performance on both specialized chemistry tasks and general chemistry questions. Surprisingly, ChemToolAgent does not consistently outperform its base LLMs without tools. Our error analysis with a chemistry expert suggests that: For specialized chemistry tasks, such as synthesis prediction, we should augment agents with specialized tools; however, for general chemistry questions like those in exams, agents' ability to reason correctly with chemistry knowledge matters more, and tool augmentation does not always help.

The concept of a Learning Health System (LHS) has been widely discussed in academic literature, yet its practical implementation remains a challenge. This paper describes the institutional journey, leadership structure, data governance policies, and technical innovations that together support a scalable and sustainable Research-Oriented LHS. Additionally, we propose an expanded data vision that aligns with interdisciplinary and translational research needs. Supplementary materials provide technical details for those interested in implementing such a model.