Renmin University of China

Two-dimensional crystals are emerging materials for nanoelectronics. Development of the field requires candidate systems with both a high carrier mobility and, in contrast to graphene, a sufficiently large electronic bandgap. Here we present a detailed theoretical investigation of the atomic and electronic structure of few-layer black phosphorus (BP) to predict its electrical and optical properties. This system has a direct bandgap, tunable from 1.51 eV for a monolayer to 0.59 eV for a five-layer sample. We predict that the mobilities are hole-dominated, rather high and highly anisotropic. The monolayer is exceptional in having an extremely high hole mobility (of order 10,000 cm2 V−1 s−1) and anomalous elastic properties which reverse the anisotropy. Light absorption spectra indicate linear dichroism between perpendicular in-plane directions, which allows optical determination of the crystalline orientation and optical activation of the anisotropic transport properties. These results make few-layer BP a promising candidate for future electronics. Two-dimensional (2D) materials with a large electronic bandgap in addition to high carrier mobility are required for future nanoelectronics. Here, the authors present a theoretical investigation of black phosphorous, a new category of 2D semiconductor with high potential for nanoelectronic applications.

Abstract Autonomous agents have long been a research focus in academic and industry communities. Previous research often focuses on training agents with limited knowledge within isolated environments, which diverges significantly from human learning processes, and makes the agents hard to achieve human-like decisions. Recently, through the acquisition of vast amounts of Web knowledge, large language models (LLMs) have shown potential in human-level intelligence, leading to a surge in research on LLM-based autonomous agents. In this paper, we present a comprehensive survey of these studies, delivering a systematic review of LLM-based autonomous agents from a holistic perspective. We first discuss the construction of LLM-based autonomous agents, proposing a unified framework that encompasses much of previous work. Then, we present a overview of the diverse applications of LLM-based autonomous agents in social science, natural science, and engineering. Finally, we delve into the evaluation strategies commonly used for LLM-based autonomous agents. Based on the previous studies, we also present several challenges and future directions in this field.

Defects usually play an important role in tailoring various properties of two-dimensional materials. Defects in two-dimensional monolayer molybdenum disulphide may be responsible for large variation of electric and optical properties. Here we present a comprehensive joint experiment-theory investigation of point defects in monolayer molybdenum disulphide prepared by mechanical exfoliation, physical and chemical vapour deposition. Defect species are systematically identified and their concentrations determined by aberration-corrected scanning transmission electron microscopy, and also studied by ab-initio calculation. Defect density up to 3.5 × 10(13) cm(-2) is found and the dominant category of defects changes from sulphur vacancy in mechanical exfoliation and chemical vapour deposition samples to molybdenum antisite in physical vapour deposition samples. Influence of defects on electronic structure and charge-carrier mobility are predicted by calculation and observed by electric transport measurement. In light of these results, the growth of ultra-high-quality monolayer molybdenum disulphide appears a primary task for the community pursuing high-performance electronic devices.

Abstract The rapid evolution of large language models (LLMs) has driven a transformative shift in artificial intelligence (AI), reshaping both research paradigms and practical applications. Distinguished from their predecessors by unprecedented scale and advanced capabilities, LLMs necessitate new frameworks for understanding their development, behavior, and societal impact. This survey systematically reviews recent advancements in LLM techniques across four key dimensions: (1) pre-training methodologies, which establish core model capabilities through large-scale self-supervised training, architectural innovations, and data curation strategies; (2) post-training techniques, including supervised fine-tuning and reinforcement learning, which adapt foundational models to downstream tasks and enhance their alignment and safety; (3) utilization strategies, such as in-context learning, prompt engineering, and agentic reasoning, that optimize real-world deployment and enable effective interaction with external environments; and (4) evaluation methods, encompassing benchmarks for key ability dimensions such as core language capabilities, reasoning, and safety, which support comprehensive and reliable assessment of model performance. Additionally, we identify critical research issues, including those concerning theoretical foundations, efficient scaling, alignment, and agentic capability, and highlight the open challenges they present. By synthesizing state-of-the-art insights and emerging trends, this survey aims to provide a systematic and comprehensive framework for understanding the trajectory, current limitations, and future directions of LLM progress.

Abstract This research investigates how entrepreneurs of small and medium enterprises (SMEs) with inadequate capabilities and limited resources drove digital transformation in their companies, a phenomenon that remains under‐researched in the extant literature. We conduct qualitative research on digital transformation to cross‐border e‐commerce undergone by 7 SMEs on the Alibaba digital platform. We inductively derive a process model that aims to describe and explain how SME entrepreneurs, with support from the digital platform service provider, drive digital transformation through managerial cognition renewal, managerial social capital development, business team building, and organizational capability building. This model expands our understanding of both digital entrepreneurship and digital transformation. It also presents new insights into how digital platform service providers can help SMEs transform and compete.

The authors found that, concurrent with the rapidly growing index investment in commodity markets since the early 2000s, prices of non-energy commodity futures in the United States have become increasingly correlated with oil prices; this trend has been significantly more pronounced for commodities in two popular commodity indices. This finding reflects the financialization of the commodity markets and helps explain the large increase in the price volatility of non-energy commodities around 2008.

The form of energy termed heat that typically derives from lattice vibrations, i.e., phonons, is usually considered as waste energy and, moreover, deleterious to information processing. However, in this Colloquium, an attempt is made to rebut this common view: By use of tailored models it is demonstrated that phonons can be manipulated similarly to electrons and photons, thus enabling controlled heat transport. Moreover, it is explained that phonons can be put to beneficial use to carry and process information. In the first part ways are presented to control heat transport and to process information for physical systems which are driven by a temperature bias. In particular, a toolkit of familiar electronic analogs for use of phononics is put forward, i.e., phononic devices are described which act as thermal diodes, thermal transistors, thermal logic gates, and thermal memories. These concepts are then put to work to transport, control, and rectify heat in physically realistic nanosystems by devising practical designs of hybrid nanostructures that permit the operation of functional phononic devices; the first experimental realizations are also reported. Next, richer possibilities to manipulate heat flow by use of time-varying thermal bath temperatures or various other external fields are discussed. These give rise to many intriguing phononic nonequilibrium phenomena such as, for example, the directed shuttling of heat, geometrical phase-induced heat pumping, or the phonon Hall effect, which may all find their way into operation with electronic analogs.

Globally increasing energy demands and environmental concerns related to the use of fossil fuels have stimulated extensive research to identify new energy systems and economies that are sustainable, clean, low cost, and environmentally benign. Hydrogen generation from solar-driven water splitting is a promising strategy to store solar energy in chemical bonds. The subsequent combustion of hydrogen in fuel cells produces electric energy, and the only exhaust is water. These two reactions compose an ideal process to provide clean and sustainable energy. In such a process, a hydrogen evolution reaction (HER), an oxygen evolution reaction (OER) during water splitting, and an oxygen reduction reaction (ORR) as a fuel cell cathodic reaction are key steps that affect the efficiency of the overall energy conversion. Catalysts play key roles in this process by improving the kinetics of these reactions. Porphyrin-based and corrole-based systems are versatile and can efficiently catalyze the ORR, OER, and HER. Because of the significance of energy-related small molecule activation, this review covers recent progress in hydrogen evolution, oxygen evolution, and oxygen reduction reactions catalyzed by porphyrins and corroles.

Due to the flexibility in modelling data heterogeneity, heterogeneous information network (HIN) has been adopted to characterize complex and heterogeneous auxiliary data in recommender systems, calledHIN based recommendation. It is challenging to develop effective methods for HIN based recommendation in both extraction and exploitation of the information from HINs. Most of HIN based recommendation methods rely on path based similarity, which cannot fully mine latent structure features of users and items. In this paper, we propose a novel heterogeneous network embedding based approach for HIN based recommendation, called HERec. To embed HINs, we design a meta-path based random walk strategy to generate meaningful node sequences for network embedding. The learned node embeddings are first transformed by a set of fusion functions, and subsequently integrated into an extended matrix factorization (MF) model. The extended MF model together with fusion functions are jointly optimized for the rating prediction task. Extensive experiments on three real-world datasets demonstrate the effectiveness of the HERec model. Moreover, we show the capability of the HERec model for the cold-start problem, and reveal that the transformed embedding information from HINs can improve the recommendation performance.

A new type of pneumonia caused by a novel coronavirus SARS-CoV-2 outbreaks recently in China and spreads into many other countries. This disease, named as COVID-19, is similar to patients infected by SARS-CoV and MERS-CoV, and nearly 20% of patients developed severe condition. Cardiac injury is a prevalent complication of severe patients, exacerbating the disease severity in coronavirus disease 2019 (COVID-19) patients. Angiotensin-converting enzyme 2 (ACE2), the key host cellular receptor of SARS-CoV-2, has been identified in multiple organs, but its cellular distribution in human heart is not illuminated clearly. This study performed the first state-of-art single cell atlas of adult human heart, and revealed that pericytes with high expression of ACE2 might act as the target cardiac cell of SARS-CoV-2. The pericytes injury due to virus infection may result in capillary endothelial cells dysfunction, inducing microvascular dysfunction. And patients with basic heart failure disease showed increased ACE2 expression at both mRNA and protein levels, meaning that if infected by the virus these patients may have higher risk of heart attack and critically ill condition. The finding of this study explains the high rate of severe cases among COVID-19 patients with basic cardiovascular disease; and these results also perhaps provide important reference to clinical treatment of cardiac injury among severe patients infected by SARS-CoV-2.

Stable ferroelectricity with high transition temperature in nanostructures is needed for miniaturizing ferroelectric devices. Here, we report the discovery of the stable in-plane spontaneous polarization in atomic-thick tin telluride (SnTe), down to a 1-unit cell (UC) limit. The ferroelectric transition temperature T(c) of 1-UC SnTe film is greatly enhanced from the bulk value of 98 kelvin and reaches as high as 270 kelvin. Moreover, 2- to 4-UC SnTe films show robust ferroelectricity at room temperature. The interplay between semiconducting properties and ferroelectricity in this two-dimensional material may enable a wide range of applications in nonvolatile high-density memories, nanosensors, and electronics.

Deep supervised learning has achieved great success in the last decade. However, its defects of heavy dependence on manual labels and vulnerability to attacks have driven people to find other paradigms. As an alternative, self-supervised learning (SSL) attracts many researchers for its soaring performance on representation learning in the last several years. Self-supervised representation learning leverages input data itself as supervision and benefits almost all types of downstream tasks. In this survey, we take a look into new self-supervised learning methods for representation in computer vision, natural language processing, and graph learning. We comprehensively review the existing empirical methods and summarize them into three main categories according to their objectives: generative, contrastive, and generative-contrastive (adversarial). We further collect related theoretical analyses on self-supervised learning to provide deeper thoughts on why self-supervised learning works. Finally, we briefly discuss open problems and future directions for self-supervised learning. An outline slide for the survey is provided.

Consumer socialization through peer communication using social media websites has become an important marketing issue through the development and increasing popularity of social media. Guided by a socialization framework, this article investigates peer communication through social media websites; individual-level tie strength and group-level identification with the peer group as antecedents; and product attitudes and purchase decisions as outcomes. Survey data from 292 participants who engaged in peer communications about products through social media confirm that the two antecedents have positive influences on peer communication outcomes. Online consumer socialization through peer communication also affects purchasing decisions in two ways: directly (conformity with peers) and indirectly by reinforcing product involvement. In addition, consumer's need for uniqueness has a moderating effect on the influence of peer communication on product attitudes. These findings have significant theoretical and managerial implications.

Large-scale pre-trained models (PTMs) such as BERT and GPT have recently achieved great success and become a milestone in the field of artificial intelligence (AI). Owing to sophisticated pre-training objectives and huge model parameters, large-scale PTMs can effectively capture knowledge from massive labeled and unlabeled data. By storing knowledge into huge parameters and fine-tuning on specific tasks, the rich knowledge implicitly encoded in huge parameters can benefit a variety of downstream tasks, which has been extensively demonstrated via experimental verification and empirical analysis. It is now the consensus of the AI community to adopt PTMs as backbone for downstream tasks rather than learning models from scratch. In this paper, we take a deep look into the history of pre-training, especially its special relation with transfer learning and self-supervised learning, to reveal the crucial position of PTMs in the AI development spectrum. Further, we comprehensively review the latest breakthroughs of PTMs. These breakthroughs are driven by the surge of computational power and the increasing availability of data, towards four important directions: designing effective architectures, utilizing rich contexts, improving computational efficiency, and conducting interpretation and theoretical analysis. Finally, we discuss a series of open problems and research directions of PTMs, and hope our view can inspire and advance the future study of PTMs.

The thermal stabilities of 66 ionic liquids (ILs) were investigated using the thermogravimetric analysis (TGA) method. Isothermal TGA studies on the ILs showed that ILs exhibit decomposition at temperatures lower than the onset decomposition temperature ( T onset ), which is determined from ramped temperature TGA experiments. Thermal decomposition kinetics of ILs was analyzed using pseudo-zero-order rate expression and their activation energy was obtained. Parameter T 0.01/10h, the temperature at which 1% mass loss occurs in 10 h, was used to evaluate the long-term thermal stability of ILs. The thermal stability of the ILs was classified to five levels according to T onset . The ILs thermal stability is dependent on the structure of ILs, i.e., cation modification, cation and anion type. The correlations between the stability and the hydrophilicity of ILs were discussed. Finally, the thermal stabilities of acetate-based ILs, amino acid ILs, and dicyanamide ILs were analyzed.

Two-dimensional materials provide extraordinary opportunities for exploring phenomena arising in atomically thin crystals. Beginning with the first isolation of graphene, mechanical exfoliation has been a key to provide high-quality two-dimensional materials, but despite improvements it is still limited in yield, lateral size and contamination. Here we introduce a contamination-free, one-step and universal Au-assisted mechanical exfoliation method and demonstrate its effectiveness by isolating 40 types of single-crystalline monolayers, including elemental two-dimensional crystals, metal-dichalcogenides, magnets and superconductors. Most of them are of millimeter-size and high-quality, as shown by transfer-free measurements of electron microscopy, photo spectroscopies and electrical transport. Large suspended two-dimensional crystals and heterojunctions were also prepared with high-yield. Enhanced adhesion between the crystals and the substrates enables such efficient exfoliation, for which we identify a gold-assisted exfoliation method that underpins a universal route for producing large-area monolayers and thus supports studies of fundamental properties and potential application of two-dimensional materials.

Abstract This article surveys corporate governance in China, as described in a growing literature published in top journals. Unlike the classical vertical agency problems in Western countries, the dominant agency problem in China is the horizontal agency conflict between controlling and minority shareholders arising from concentrated ownership structure; thus one cannot automatically apply what is known about the USA to China. As these features are also prevalent in many other countries, insights from this survey can also be applied to countries far beyond China. We start by describing controlling shareholder and agency problems in China, and then discuss how law and institutions are particularly important for China, where controlling shareholders have great power. As state-owned enterprises have their own features, we separately discuss their corporate governance. We also briefly discuss corporate social responsibility in China. Finally, we provide an agenda for future research.

Graph representation learning has emerged as a powerful technique for addressing real-world problems. Various downstream graph learning tasks have benefited from its recent developments, such as node classification, similarity search, and graph classification. However, prior arts on graph representation learning focus on domain specific problems and train a dedicated model for each graph dataset, which is usually non-transferable to out-of-domain data. Inspired by the recent advances in pre-training from natural language processing and computer vision, we design Graph Contrastive Coding (GCC) --- a self-supervised graph neural network pre-training framework --- to capture the universal network topological properties across multiple networks. We design GCC's pre-training task as subgraph instance discrimination in and across networks and leverage contrastive learning to empower graph neural networks to learn the intrinsic and transferable structural representations. We conduct extensive experiments on three graph learning tasks and ten graph datasets. The results show that GCC pre-trained on a collection of diverse datasets can achieve competitive or better performance to its task-specific and trained-from-scratch counterparts. This suggests that the pre-training and fine-tuning paradigm presents great potential for graph representation learning.

Recently, significant progress has been made in sequential recommendation with deep learning. Existing neural sequential recommendation models usually rely on the item prediction loss to learn model parameters or data representations. However, the model trained with this loss is prone to suffer from data sparsity problem. Since it overemphasizes the final performance, the association or fusion between context data and sequence data has not been well captured and utilized for sequential recommendation.

ABSTRACT We develop FinBERT, a state‐of‐the‐art large language model that adapts to the finance domain. We show that FinBERT incorporates finance knowledge and can better summarize contextual information in financial texts. Using a sample of researcher‐labeled sentences from analyst reports, we document that FinBERT substantially outperforms the Loughran and McDonald dictionary and other machine learning algorithms, including naïve Bayes, support vector machine, random forest, convolutional neural network, and long short‐term memory, in sentiment classification. Our results show that FinBERT excels in identifying the positive or negative sentiment of sentences that other algorithms mislabel as neutral, likely because it uses contextual information in financial text. We find that FinBERT's advantage over other algorithms, and Google's original bidirectional encoder representations from transformers model, is especially salient when the training sample size is small and in texts containing financial words not frequently used in general texts. FinBERT also outperforms other models in identifying discussions related to environment, social, and governance issues. Last, we show that other approaches underestimate the textual informativeness of earnings conference calls by at least 18% compared to FinBERT. Our results have implications for academic researchers, investment professionals, and financial market regulators.