Design and validation of reinforcement learning - empowered generative AI teaching capability enhancement strategies for diverse majors in higher vocational education

This study addresses the marked differences in knowledge structures across higher vocational majors.Such disparities prevent generative artificial intelligence (GenAI) teaching support from precise major-specific matching, causing unstable improvements in teaching ability.Multi-professional adaptive generative feedback learning (MPA-GRL) model is proposed.By constructing professional embedded modulation mechanism and hierarchical RL (reinforcement learning) collaborative optimisation strategy, the dynamic adaptation and teaching feedback regulation of GenAI in different professional situations are realised.Results indicated that after a four-month intervention, the experimental group's average teaching ability rose from baseline to 3.90-4.28,an increase of 0.82-0.97points, compared with only 0.13-0.30points in the control group.Between-group differences became statistically significant at month four, with p < 0.005 for most majors.The improvement range showed no significant disparity across majors (F = 0.84, p = 0.652), confirming favourable cross-disciplinary stability of the model.