Peng Liu

Existing fine-tuning data selection methods for large language models (LLMs) often struggle to balance two aspects: the evaluation of legal fine-tuning data quality and the maintenance of syntactic diversity. The assessment of data quality consists of two dimensions: legal content quality, encompassing logical rigor, legal basis traceability, fact coverage, and actionable guidance, and instructional complexity, which measures how challenging an instruction–response pair is for the model to learn. Syntactic diversity, in contrast, captures the variety of sentence structures and dependency patterns within the dataset, reflecting how sentences differ in grammatical organization, for example through variations between active and passive voice, clause nesting, or modifier placement. To address these issues, this paper proposes LEAD (Legal Efficiency and Diversity), a fine-tuning data selection framework for legal LLMs that integrates dual-objective optimization with syntactic structure clustering. LEAD consists of two modules: the Legal Efficiency Training Score (LETS), which evaluates each sample’s legal quality, and Syntactic Diversity Clustering (SDC), which applies graph-kernel methods to ensure syntactic diversity. Experiments show that using only 5% of the original dataset selected by LEAD, the resulting model outperforms full-data training by 14.9% and achieves an average 12.3% improvement across legal benchmarks, demonstrating strong potential for efficient legal LLM fine-tuning.

BackgroundMild cognitive impairment is a common neurological disorder. Moxibustion has been shown to be effective in treating MCI, but its therapeutic mechanisms still remain unclear. This study mainly aimed to investigate the modulation effect of moxibustion treatment for patients with MCI by functional magnetic resonance imaging.MethodsA total of 47 patients with MCI and 30 healthy controls participated in resting-state fMRI imaging scans. Patients with MCI were randomly divided into true moxibustion group and sham moxibustion group. The degree centrality approach was applied to distinguish altered brain functions. Correlation analysis was then performed to examine the relationships between the neuroimaging findings and clinical symptoms.ResultsCompared with HCs, patients with MCI mainly showed decreased DC in the left middle frontal cortex and bilateral middle cingulate cortex. After moxibustion treatment, the SHAM group had no significant DC findings, while TRUE group mainly showed significant increased DC in the bilateral MFC and MCC, as well as decreased DC in the left middle occipital cortex. Repeated measures analysis of variance showed significant interactions between the two groups of patients with MCI. In addition, the higher Mini-Mental State Examination score was significantly positively correlated with increased DC in the right MFC and left MCC after moxibustion treatment.ConclusionOur findings demonstrate that the potential value of moxibustion treatment on MCI, which adds new insights into the popular view that moxibustion treatment may slow cognitive decline in patients with MCI.

Over the past two decades, hundreds of articles have investigated the mechanisms underlying the phenomenon that masked stimuli reduce people's response performance to subsequent compatible stimuli, for example, the negative compatibility effect (NCE). Whether the NCE results from motor inhibition, object updating, or both is still being debated. We used the digital masked prime task for 3 consecutive days to strengthen stimulus‐response associations in relevant and irrelevant contexts (whether the mask consisted of task‐relevant features or not) and employed response time distribution analysis to investigate the contributions of motor inhibition and object updating to the NCE. The results showed that the NCE appeared in the irrelevant condition on days 2 and 3, and it increased with response latency on day 3. In contrast, in the relevant condition, the NCE occurred regardless of test day or response latency, and was unaffected by either. These different patterns of results indicated that the cause of the NCE was different in the relevant and irrelevant conditions. In the relevant condition, the results suggested that the NCE was solely due to object updating, whereas in the irrelevant condition, the results indicated that the NCE was solely due to motor inhibition. This study reconciled the previous debate and revealed the mechanisms by which unconscious information influences behavioral performance in different contexts.

In this paper, we establish a mathematical duality between utility transforms and probability distortions. These transforms play a central role in decision under risk by forming the foundation for the classic theories of expected utility, dual utility, and rank-dependent utility. Our main results establish that probability distortions are characterized by commutation with utility transforms, and utility transforms are characterized by commutation with probability distortions. These results require no additional conditions, and hence each class can be axiomatized with only one property. Moreover, under monotonicity, rank-dependent utility transforms can be characterized by set commutation with either utility transforms or probability distortions.

Lifelong premature ejaculation (LPE) is a common male sexual dysfunction. Lack of active control for rapid ejaculation brought great distress to sexual harmony, and even fertility. Previous neurophysiology studies revealed an ejaculation-related control mechanism in the brain. However, it remains unclear whether this inhibitory network is altered in LPE patients. The present study investigated the central inhibitory network function of LPE patients by using stop signal task-related functional magnetic resonance imaging and resting-state functional connectivity analysis. The results showed no difference in task-related behavioral performance or neural activation during response inhibition between LPE patients and controls. However, LPE patients showed a significantly different correlation pattern between the stop signal reaction time (SSRT) and left inferior frontal gyrus activation during successful inhibition, in which a typical negative correlation between SSRT and the activation was completely disappeared in patients. In addition, using the left inferior frontal gyrus as a seed, patients showed weaker functional connectivity between the seed and two areas (left dental nucleus and right frontal pole) compared with controls. These data suggest that LPE patients have an abnormal brain control network, which may contribute to the reduced central control of rapid ejaculation. This study provides new insights into the neural mechanism of LPE involving the central inhibitory network, which may offer an underlying intervention target for future treatment.

A human driver and an automated driving system might share control of automated vehicles in the near future. This raises many concerns associated with the assignment of responsibility for negative outcomes caused by them; one is that the human driver might be required to bear the brunt of moral and legal responsibilities. The psychological consequences of responsibility misattribution have not yet been examined. We designed a hypothetical crash similar to Uber’s 2018 fatal crash. We incorporated five legal responsibility attributions. Participants chose their preferred liability attribution and then were randomly assigned into one of the five actual liability attribution conditions. They then responded to a series of questions concerning liability assignment, the crash, and AVs. Slightly more than 50% of participants thought that the human driver should bear full or primary liability. Legal responsibility misattribution negatively influenced these mentioned responses, regardless of overly attributing human or manufacturer liability. Overly attributing human liability had more negative influences. Improper liability attribution might hinder the adoption of AVs. Public opinion should not be ignored in developing a legal framework for AVs.

Unlike eggs and diatoms, most single cells in nature do not have structured shells to provide extensive protection. It is a challenge to artificially confer shell structures on living cells to improve their inherent properties and functions. We discuss four different types of cellular shellizations: man‐made hydrogels, sol‐gels, polyelectrolytes, and mineral shells. We also explore potential applications, such as cell storage, protection, delivery, and therapy. We suggest that shellization could provide another means to regulate and functionalize cells. Specifically, the integration of living cells and non‐living functional shells may be developed as a novel strategy to create “super” or intelligent cells. Unlike biological approaches, this material‐based bio‐interface regulation is inexpensive, effective, and convenient, opening up a novel avenue for cell‐based technologies and practices.

The present study critically evaluates whether school leadership influences student learning homogenously regardless of school contexts. It examined relationships between four principal leadership variables (envisioning, instructional management, promoting professional development, empowerment) and two types of student outcomes (enjoyment in learning science, science achievement,) in different school contexts (in terms of the availability of science resources, quality of science teachers, and school socioeconomic status (SES)). The sample comprised 248,620 students and 9,370 principals in 35 developed countries who participated in the Programme for International Student Assessment (PISA) 2015. Latent profile analysis (LPA) showed that schools operated in three types of school contexts with different levels of science resources, proportion of quality science teachers, and school SES. There were also differences in the pattern of leadership practices across the three types of school contexts. Three-level hierarchical linear modeling (HLM) showed that among the four leadership variables, only instructional management was positively associated with students’ enjoyment of science in schools with less science resources and quality science teachers. Therefore, instructional management had compensatory effects for students in less-endowed schools. In contrast, principal leadership related to envisioning, teacher professional development, and empowerment was not positively related to students’ science learning in all three school contexts.