Shu Su

As Large Language Models (LLMs) serve a global audience, alignment must transition from enforcing universal consensus to respecting cultural pluralism. We demonstrate that dense models, when forced to fit conflicting value distributions, suffer from \textbf{Mean Collapse}, converging to a generic average that fails to represent diverse groups. We attribute this to \textbf{Cultural Sparsity}, where gradient interference prevents dense parameters from spanning distinct cultural modes. To resolve this, we propose \textbf{\textsc{CuMA}} (\textbf{Cu}ltural \textbf{M}ixture of \textbf{A}dapters), a framework that frames alignment as a \textbf{conditional capacity separation} problem. By incorporating demographic-aware routing, \textsc{CuMA} internalizes a \textit{Latent Cultural Topology} to explicitly disentangle conflicting gradients into specialized expert subspaces. Extensive evaluations on WorldValuesBench, Community Alignment, and PRISM demonstrate that \textsc{CuMA} achieves state-of-the-art performance, significantly outperforming both dense baselines and semantic-only MoEs. Crucially, our analysis confirms that \textsc{CuMA} effectively mitigates mean collapse, preserving cultural diversity. Our code is available at https://github.com/Throll/CuMA.

The building industry has significant environmental, economic, and social impacts. The trend of construction industrialization to promote sustainable development is becoming increasingly evident. As an essential component of this process, prefabricated components provide a foundation for studying the sustainability of prefabricated buildings. This study proposes a life cycle sustainability assessment (LCSA) model involving environmental, economic, and social aspects to assess the sustainability of prefabricated components. The sustainability impacts on the raw material extraction and production, off-plant transport, material processing and component manufacturing, on-plant transport, and component storage stages are comprehensively assessed. The LCSA model is applied to four types of prefabricated components: interior wallboards, precast stairs, composite beams, and composite floor slabs. The results show that the precast stairs exhibit the highest sustainability score and that the performances of environmental, economic, and social pillars of four components are slightly different. These findings enhance our understanding of the sustainability of prefabricated components and broaden the scope of LCSA applications. The established sustainability assessment model is expected to help guide prefabrication scheme designs and production plan optimization, further encouraging the wider adoption of sustainable practices in construction.

Residential districts, the primary locations for older adults, struggle to satisfy the needs of the rising older population due to ageing infrastructure, etc. China has begun to focus on this issue through the popularization of approaches such as active ageing theory (AAT), and has implemented ageing suitability renovations. Meanwhile, recent evaluation models failed to validate their logic and standardization. The current evaluation models prioritize the indoor environment above the external environment. Therefore, this study attempts to provide a comprehensive evaluation model of the ageing suitability of residential districts. First, 23 indicators of ageing suitability were identified by a systematic literature review and expert interviews; next, the indicators were categorized using the active ageing theory. Second, the grey correlation analysis approach was used to calculate the weights of evaluation indicators, and the existing evaluation criteria were incorporated to formulate the evaluation criteria for the project. Eventually, the evaluation model was constructed. Finally, the feasibility of the model was validated by a case study. The findings of this study can help provide directions for the future renovation of old residential districts (for example, the activity area should be over 200 m²) and the construction of suitable houses for older adults, such as the installation of a continuous handrail, to alleviate the increasingly grave problem of ageing and promote the development of the pension industry.

Life cycle assessment (LCA) is a widely-used international environmental evaluation and management method. However, the conventional LCA is in a static context without temporal and spatial variations considered, which fails to bring accurate evaluation values and hinders practical applications. Dynamic LCA research has developed vigorously in the past decade and become a hot topic. However, systematical analysis of spatiotemporal dynamic variations and comprehensive operable dynamic models are still lacking. This study follows LCA paradigm and incorporates time- and space-dependent variations to establish a spatiotemporal dynamic LCA model. The dynamic changes are classified into four types: dynamic foreground elementary flows, dynamic background system, dynamic characterization factors, and dynamic weighting factors. Their potential dynamics and possible quantification methods are analyzed. The dynamic LCA model is applied to a residential building, and significant differences can be observed between dynamic and static assessment results from both temporal and spatial perspectives. This study makes a theoretical contribution by establishing a comprehensive dynamic model with both temporal and spatial variations involved. It is expected to provide practical values for LCA practitioners and help with decision-making and environmental management.

With the rapid development of information technology, building information modeling (BIM) has been widely applied in green building evaluation, attracting extensive attention. It is necessary to systematically summarize the research status and explore potential trends of BIM-based green building assessment research. This study reviewed 239 research publications using the Web of Science core collection database and adopted CiteSpace to conduct a quantitative literature review. This review presents the geographical distribution of related studies, conducts co-citation analyses for journals, authors, and documents, identifies high-frequency keywords, and explores emerging research frontiers. A knowledge graph was established based on the outcomes of this review to present the development process of this research field. This study may assist researchers and practitioners to better understand the status and trends in BIM-based green building assessment research, guiding the research direction for future studies.

Life cycle assessment (LCA) is a comprehensive and important environmental management tool around the world. However, lacking temporal information has been a major challenge. In the past decade, dynamic LCA (DLCA), which incorporates temporal variations into assessment, has been an emerging research topic with increasing publications. A timely comprehensive review is needed to present current progress and discuss future directions. This article reviews 144 DLCA articles quantitatively and qualitatively. A bibliometric approach is adopted to conduct co-occurrence analysis and cluster analysis of DLCA studies. The research progress, approaches, and limitations of three temporal variation types (i.e., dynamic life cycle inventory, dynamic characterization factors, and dynamic weighting factors) in DLCA studies are systematically analyzed and discussed. It is concluded that: (1) dynamic inventory analysis is usually conducted by collecting time-differentiated data at each time step. Field monitoring, simulation, scenario analysis, and prediction based on historical data are common approaches. (2) Dynamic characterization studies primarily focus on two impact categories: global warming and toxicity. More studies are in need. (3) Various methods and indicators (i.e., dynamic pollution damage cost, temporal environmental policy targets, and discount rates) are used to solve the dynamic weighting issue, and they have specific limitations. Finally, three interesting topics are discussed: comparison between dynamic and static results, the large data amount issue, and the trend of tools development. This review offers a holistic view on temporal variations in DLCA studies and provides reference and directions for future dynamic studies.