Blended learning has emerged as a popular trend in education; however, as students from higher vocational colleges who are often in the lower 50% of national standardized entrance exam achievers, their teachers face unique challenges in implementing blended learning. This study summarized effective blended learning strategies in higher vocational education and training (VET) classrooms by comparing 53 award-winning and 40 daily course videos in China. Firstly, video analysis and Lag sequential analysis were employed to identify effective strategies for implementing blended learning in diverse VET course types. A set of general and specific blended learning strategies was developed to help VET teachers adapt instructional approaches accordingly. Secondly, a questionnaire survey among 215 VET teachers revealed positive perceptions of the strategies in terms of usability, ease of use, perceived behavioral control, and intention to use them. The present research provides valuable guidance for VET teachers to effectively implement blended learning strategies in diverse course types, contributing to the understanding of effective blended learning strategies in VET and addressing the gap in research for this unique teaching stage.

This study examined how different forms of family cultural capital are linked to Chinese graduate students’ academic engagement and research self-efficacy, and whether these pathways differ by rural–urban background. Drawing on cultural capital theory and social cognitive theory, the study distinguished embodied, objectified, and institutionalized family cultural capital and conceptualized academic engagement as a behavioral pathway through which family-based resources may become relevant to research. Survey data were collected from 412 graduate students at five public universities in China. Partial least squares structural equation modeling with 5000 bootstrap resamples was used to test direct and indirect pathways separately for students from rural and urban backgrounds. The results showed that embodied family cultural capital was positively related to academic engagement in both groups, and academic engagement was positively related to research self-efficacy. Academic engagement mediated the relationship between embodied family cultural capital and research self-efficacy for both rural- and urban-background students. However, embodied family cultural capital only had a significant direct relationship with research self-efficacy among urban-background students. Objectified family cultural capital was only negatively related to research self-efficacy among rural-background students, while institutionalized family cultural capital showed no significant relationships in either group. These findings suggest that family cultural capital is not uniformly converted into research self-efficacy in graduate education. Its role depends on the form of cultural capital and the rural–urban context in which academic engagement is enacted. Theoretically, this study shows that the conversion of family cultural capital into research self-efficacy is form-specific and context-dependent. Practically, it identifies academic engagement as a key leverage point for promoting equitable research development among graduate students from diverse family backgrounds.

Research self-efficacy (RSE) is critical for graduate students. However, how emerging AI technologies shape RSE remains unclear. Grounded in the Technology Readiness Index (TRI) and Social Cognitive Theory, this study examined how AI technology readiness, AI trust, and supervisor academic support are associated with RSE. The study also examined whether these associations differ across disciplinary groupings. Cross-sectional data were collected from 425 master’s students at 19 Chinese universities. All participants self-reported regular use of AI in their research activities. The data were analyzed using partial least squares structural equation modeling (PLS-SEM) with multigroup analysis. Readiness facilitators were positively associated with AI trust and with perceived supervisor academic support. Readiness inhibitors were also positively associated with AI trust. Supervisor academic support was significantly associated with RSE. The direct association between readiness and RSE was not statistically significant. However, statistically significant indirect associations were observed via supervisor academic support and AI trust. Importantly, partial measurement invariance was established across groups. No between-group differences in path coefficients reached statistical significance. This indicates structural equivalence across the humanities and social sciences grouping and the science, engineering, and medicine grouping. The findings offer a preliminary extension of the TRI from adoption outcomes to efficacy beliefs in research training. They also inform tentative, discipline-uniform strategies for AI literacy education and supervisory support among master’s students who regularly use AI for research purposes.