1
Rethinking Scaffolding in LLM Tutors: The Interactional Mismatch Between Benchmarks and Real-World Deployments
揭示LLM导师中脚手架评估的隐含假设漏洞，指出基准测试与真实课堂场景的交互错位。
arXiv:2606.15766v1 Announce Type: new Abstract: A central pedagogical value evaluated in AI tutor benchmarks is scaffolding: guiding students through …