<p>In higher education, evaluating students’ academic papers remains a complex and time-consuming task, particularly under increasing pressures on educators’ time and cognitive capacity. This study investigates the collaborative potential between generative artificial intelligence (GAI) and teachers in academic paper evaluation. A total of 60 undergraduate theses were assessed under two conditions: teacher-only mode and teacher-GAI collaborative mode. Additionally, semi-structured interviews were conducted with 15 participating teachers and 8 students whose papers had been evaluated. Quantitative analyses (Cohen’s Kappa and Mann–Whitney U tests) were conducted to examine consistency and differences between the two evaluation modes, while qualitative analyses of teacher and student interviews explored perceptions of feedback quality and role distribution. Results showed that the grading results under the teacher-GAI collaborative mode demonstrated moderate consistency with those from teacher-only evaluations, while its evaluative comments were richer in content and clearer in expression. Thematic analysis further identified four core roles played by teachers (AI Orchestrator, Scholarly Mentor, Evaluation Manager, and Personalized Evaluator) and four by GAI (Preliminary Manuscript Assessor, Information Retrieval Assistant, Text Quality Reviewer, and Report Writing Assistant). Based on these findings, an Integrated Teacher-GAI Academic Evaluation Framework was proposed, outlining stage-specific role distribution and collaborative strategies. This study offers both theoretical grounding and practical implications for advancing human-AI collaboration in academic assessment within higher education.</p>