<br> Despite rising concerns about sycophancy—excessive agreement or flattery from artificial intelligence (AI) systems—little is known about its prevalence or consequences. We show that sycophancy is widespread and harmful. Across 11 state-of-the-art models, AI affirmed users’ actions 49% more often than humans, even when queries involved deception, illegality, or other harms. In three preregistered experiments (<br> N</i><br> = 2405), even a single interaction with sycophantic AI reduced participants’ willingness to take responsibility and repair interpersonal conflicts, while increasing their conviction that they were right. Despite distorting judgment, sycophantic models were trusted and preferred. This creates perverse incentives for sycophancy to persist: The very feature that causes harm also drives engagement. Our findings underscore the need for design, evaluation, and accountability mechanisms to protect user well-being.<br> <<<

<br> Background<br> Insomnia is one of the most common non-motor comorbidities of Parkinson’s disease (PD) and often before the onset of motor symptoms. Identifying the molecular mechanisms of insomnia may facilitate the early diagnosis of PD and contribute to therapeutic development.<br> <br> <br> Methods<br> <br> Five human PD substantia nigra (SN) bulk-seq datasets (GSE20141, GSE7621, GSE20164, GSE20163, and GSE20333), with an insomnia-related gene list, were acquired from GEO and Genecard databases. First, the integration of GSE20141 and GSE7621 was analyzed to identify insomnia-related DEGs using limma and the WGCNA framework. GSE20164 and GSE20163 combination were used as a training set for insomnia-related hub gene recognition. Furthermore, the aforementioned four datasets, along with an independent validation set (GSE20333), were cross-validated for insomnia-related diagnostic model construction. The human PD-SN single-cell profile (GSE140231) was utilized for exploring the mechanisms underlying the heterogeneity of insomnia-related hub genes in spatial and temporal contexts. Furthermore, a cutting-edge artificial intelligence (AI)-driven framework (DrugRefLector) and molecular docking techniques was used to identify an optimal agent for the treatment of PD based on the GSE20164 and GSE20163 integrated dataset. Finally, an<br> <i>in vitro</i><br> q-RT-PCR experiment was conducted to estimate the targeted gene expression.<br> <br> <br> <br> Results<br> TAZ (WWTR1) is associated with the increased expression of insomnia-related diagnostic markers linked to PD pathogenesis, mainly in neurons, and has excellent predictive performance for PD diagnosis. Furthermore, BRD-K97481123 can be considered as a potential therapeutic agent for the treatment of PD by targeting TAZ.<br> <br> <br> Conclusion<br> By integrating AI pipelines and multi-omics, our study first traced TAZ mechanisms in PD pathogenesis and elaborated on TAZ’s predictive and druggable potential for PD patients.<br> <<<

<br> Sequential movements rely on two information sources: external sensory cues and internal memory representations. Although often both sources jointly drive sequential behavior, previous research has primarily examined them in isolation. To address this, we trained participants (<br> <i>n</i><br> = 26, 15F) to perform sequences of rapid finger presses in response to numerical cues. Sensory influence was measured by varying the number of visible cues, and memory influence was determined by comparing repeating and random sequences. Early in learning, participants integrated sensory and memory information: repeating sequences were performed more quickly when more cues were visible. After learning, when repeating sequences were predictable with certainty, participants relied solely on memory and ignored sensory cues. However, when this certainty was manipulated by introducing occasional violations within repeating sequences, participants reverted to integrating memory with sensory cues. We propose a computational model that successfully predicted both speed and accuracy of individual presses. Critically, this model relied on the assumption that multiple movements are planned independently of each other. This independence assumption was then validated by examining response patterns to isolated violations in repeating sequences. Finally, we provide evidence into how sequence memories can be flexibly deactivated and reactivated in response to these violations. Together, these results reveal how brain dynamically integrates sensory and memory information to produce sequences of movements.<br> <<<

Abstract<br> <br> Millions of human genomes have been genotyped by national biobanks worldwide. Training large language models (LLM) with this data may lead to a universal model of human genome with tremendous potential. Yet the quadrillions (10<br> 15<br> ) of nucleotides— resulting from genome length multiplied by population size—pose formidable challenges for modeling. In this study, we propose a novel AI framework designed to scale with this data and support diverse analytical tasks. To demonstrate this scheme, we developed SNPBag—a foundation model focusing on single nucleotide polymorphism (SNP). With 0.8 billion parameters, it is trained on one million synthesized human genomes, corresponding to a total of 6 trillion SNP tokens. SNPBag showed superior performance in benchmarking of multiple tasks. In genotype imputation, it achieves state-of-the-art (SOTA) accuracy. In haplotype phasing, it rivals the best method with a 72-fold speedup. By encoding 6 million SNPs per genome into a 0.75 MB embedding, SNPBag enables efficient storage, transfer and downstream applications. In particular, the genome embeddings facilitate rapid ancestry inference across global populations and detection of genetic relationships up to 12th-degree relatives. Collectively, SNPBag introduces a new paradigm for scalable, unified and multitask analysis of the ever-growing human variation data.<br> <<<

Since pioneering work of Hinton et al., knowledge distillation based on Kullback-Leibler Divergence (KL-Div) has been predominant, and recently its variants have achieved compelling performance. However, KL-Div only compares probabilities of the corresponding category between the teacher and student while lacking a mechanism for cross-category comparison. Besides, KL-Div is problematic when applied to intermediate layers, as it cannot handle non-overlapping distributions and is unaware of geometry of the underlying manifold. To address these downsides, we propose a methodology of Wasserstein Distance (WD) based knowledge distillation. Specifically, we propose a logit distillation method called WKD-L based on discrete WD, which performs cross-category comparison of probabilities and thus can explicitly leverage rich interrelations among categories. Moreover, we introduce a feature distillation method called WKD-F, which uses a parametric method for modeling feature distributions and adopts continuous WD for transferring knowledge from intermediate layers. Comprehensive evaluations on image classification and object detection have shown (1) for logit distillation WKD-L outperforms very strong KL-Div variants; (2) for feature distillation WKD-F is superior to the KL-Div counterparts and state-of-the-art competitors. The source code is available at https://peihuali.org/WKD <<<

The discovery of circulating fetal and tumor cell-free DNA (cfDNA) molecules in plasma has opened up tremendous opportunities in noninvasive diagnostics such as the detection of fetal chromosomal aneuploidies and cancers and in posttransplantation monitoring. The advent of high-throughput sequencing technologies makes it possible to scrutinize the characteristics of cfDNA molecules, opening up the fields of cfDNA genetics, epigenetics, transcriptomics, and fragmentomics, providing a plethora of biomarkers. Machine learning (ML) and/or artificial intelligence (AI) technologies that are known for their ability to integrate high-dimensional features have recently been applied to the field of liquid biopsy. In this review, we highlight various AI and ML approaches in cfDNA-based diagnostics. We first introduce the biology of cell-free DNA and basic concepts of ML and AI technologies. We then discuss selected examples of ML- or AI-based applications in noninvasive prenatal testing and cancer liquid biopsy. These applications include the deduction of fetal DNA fraction, plasma DNA tissue mapping, and cancer detection and localization. Finally, we offer perspectives on the future direction of using ML and AI technologies to leverage cfDNA fragmentation patterns in terms of methylomic and transcriptional investigations. <<<

Abstract<br> <br> The automation of science is a long-standing ambition in artificial intelligence (AI) research<br> 1,2<br> . Although the community has made substantial progress in automating individual components of the scientific process, a system that autonomously navigates the entire research life cycle—from conception to publication—has remained out of reach. Here we present a pipeline for automating the entire scientific process end to end. We present The AI Scientist, which creates research ideas, writes code, runs experiments, plots and analyses data, writes the entire scientific manuscript, and performs its own peer review. Its ideas, execution and presentation are of sufficient quality that the manuscript generated by this AI system passed the first round of peer review for a workshop of a top-tier machine learning conference. The workshop had an acceptance rate of 70%. Our system leverages modern foundation models<br> 3–5<br> within a complex agentic system. We evaluate The AI Scientist in two settings: a focused mode using human-provided code templates as an initial scaffold for conducting research on a specific topic and a template-free, open-ended mode that leverages agentic search for wider scientific exploration<br> 6,7<br> . Both settings produce diverse ideas and automatically test, report on and evaluate them. This achievement demonstrates the growing capacity of AI for making scientific contributions and signifies a potential paradigm shift in how research is conducted. As with any impactful new technology, there could be important risks, including taxing overwhelmed review systems and adding noise to the scientific literature. However, if developed responsibly, such autonomous systems could greatly accelerate scientific discovery.<br> <<<

AbstractAstrocytes are a direct target of neuromodulators and can influence neuronal activity on broad spatial and temporal scales in response to a rise in cytosolic calcium. However, our knowledge about how astrocytes are recruited during different animal behaviors remains limited. To measure astrocyte activity calcium in vivo during normative behaviors, we utilize a high-resolution, long working distance multicore fiber optic imaging system that allows visualization of individual astrocyte calcium transients in the cerebral cortex of freely moving mice. We define the spatiotemporal dynamics of astrocyte calcium changes during diverse behaviors, ranging from sleep-wake cycles to the exploration of novel objects, showing that their activity is more variable and less synchronous than apparent in head-immobilized imaging conditions. In accordance with their molecular diversity, individual astrocytes often exhibit distinct thresholds and activity patterns during explorative behaviors, allowing temporal encoding across the astrocyte network. Astrocyte calcium events were induced by noradrenergic and cholinergic systems and modulated by internal state. The distinct activity patterns exhibited by astrocytes provides a means to vary their neuromodulatory influence in different behavioral contexts and internal states. <<<

Abstract<br> Biomolecular functions are governed by dynamic conformational ensembles rather than static structures. While models like AlphaFold have revolutionized static structure prediction, accurately capturing the equilibrium distribution of all-atom biomolecular interactions remains a significant challenge due to the high computational cost of molecular dynamics (MD). We present AnewSampling, a transferable generative foundation framework designed for the high-fidelity sampling of all-atom equilibrium distributions, which is the first model to faithfully reproduce MD at the all-atom level. It uses a novel quotient-space generative framework to ensure mathematical consistency and leverages the largest self-curated database of protein-ligand trajectories to date, with over 15 million conformations. Statistically, AnewSampling consistently outperforms all prior generative methods on the ATLAS monomer benchmark, and the all-atom capabilities of AnewSampling enable close statistical alignment with ground-truth MD for evaluating atomic biomolecular interactions in protein-ligand dynamics. Furthermore, AnewSampling successfully recovers coupled ligand and side-chain motions in CDK2 systems, overcoming a major sampling hurdle inherent to conventional MD. AnewSampling enables rapid exploration of conformational landscapes prior to intensive simulations, elucidating fundamental biophysical mechanisms and accelerating the broader design of functional biomolecules. <<<