Purpose <p>To evaluate the Cooperative Training Community (CTC) framework for medical undergraduate research training using a mixed-methods design, and to examine the extent to which observed group differences in research outcomes are attributable to the CTC intervention versus confounding by gender and academic year.</p> Method <p>This retrospective mixed-methods study was conducted at Xiangya School of Medicine, Central South University, China using data collected in April 2022. Qualitative data comprised transcripts from a semi-structured focus group with one CTC team. Quantitative data were drawn from surveys completed by 295 undergraduates (64 CTC, 231 non-CTC) and 207 graduates (17 CTC, 190 non-CTC), assessing research engagement, self-efficacy, creative efficacy, and innovation performance. Unadjusted group comparisons were followed by multivariate logistic regression adjusting for gender and academic year. A pre-specified sensitivity analysis was conducted in the sophomore subgroup (<i>n</i> = 88), the only academic year stratum in which gender distribution did not differ significantly between CTC and non-CTC students (<i>p</i> = 0.129), allowing a more controlled within-stratum comparison.</p> Results <p>Unadjusted analyses suggested that CTC undergraduates outperformed non-CTC peers on multiple research engagement and productivity outcomes. However, CTC and non-CTC groups differed significantly in both gender distribution (female: 73.4% vs. 53.2%, <i>p</i> = 0.004) and academic year (<i>p</i> &lt; 0.001). After multivariate adjustment, academic year emerged as the most consistent independent predictor of research productivity across all five outcomes examined (all <i>p</i> ≤ 0.019). Female gender independently predicted paper publication (OR = 2.503, <i>p</i> = 0.009) and award attainment (OR = 2.411, <i>p</i> = 0.003). CTC group membership was not an independent predictor of any productivity outcome after adjustment, with the exception of project participation—a structural feature of CTC membership rather than an educational outcome. In the sophomore sensitivity analysis, no significant CTC-related differences were found in any behavioural or productivity measure; however, CTC students reported significantly lower research self-efficacy (mean difference: −17.81, <i>p</i> = 0.005, d = 0.639) and creative efficacy (<i>p</i> = 0.046, d = 0.449), with the largest deficits in items measuring autonomous higher-order skills. Qualitative findings were uniformly positive, consistent with social desirability effects in a non-anonymous focus group setting. Among graduate students, the small CTC subsample (<i>n</i> = 17) precluded definitive conclusions; however, a gender-stratified sensitivity analysis found that CTC graduates reported stronger perceptions of research-thesis disconnection than non-CTC peers in both male and female subgroups (Δ ≈ 0.45–0.52), suggesting role strain independent of the group’s gender imbalance.</p> Conclusion <p>In this single-institution study, the apparent advantages of CTC participation were largely attributable to confounding by gender and academic year rather than the framework’s educational effects. Academic year was the strongest independent predictor of research productivity, underscoring the primacy of accumulated research exposure over training model. CTC participation was associated with lower self-efficacy for autonomous research skills among year-matched undergraduates, consistent with theoretical predictions regarding over-scaffolding. These findings highlight the methodological importance of adjusting for demographic confounders in educational intervention evaluations, and suggest that future iterations of structured collaborative training frameworks should incorporate deliberate mechanisms to foster learner autonomy alongside participation.</p>