<p>This study investigates how different formative–summative assessment weight configurations affect evaluation validity and discriminatory power in a landscape architecture planning and design course, in the context of contemporary undergraduate education evaluation reforms in China. A randomized controlled experiment with 90 landscape architecture students at a provincial university in eastern China implemented three weight allocation schemes: 30%:70%, 50%:50%, and 70%:30% (formative: summative). Validity analysis, discrimination analysis, and curve fitting were used to examine the effects of weight configuration on assessment quality. The findings indicate that (1) weight configuration has a pronounced nonlinear effect on evaluation validity: the balanced configuration (50%:50%) yielded higher validity (r = 0.845) than both the traditional (30%:70%, r = 0.723) and process-oriented configurations (70%:30%, r = 0.776), with validity approaching its optimum near a formative weight of approximately 53%; (2) the effect of weight configuration on discriminatory power exhibits diminishing marginal returns: the process-oriented configuration produced the highest discrimination index (D = 0.435), followed by the balanced configuration (D = 0.412), both substantially exceeding the traditional configuration (D = 0.325)—when validity and discriminatory power are considered jointly, a formative weight of approximately 55% yielded stronger overall performance; and (3) different competency dimensions respond differentially to assessment approaches—site analysis and design thinking showed higher validity under formative assessment, whereas spatial composition and communication skills were more effectively captured by summative assessment. This study provides empirical evidence and an adaptable methodological framework for evidence-based refinement of assessment systems in design courses.</p>