A new paper details the Learning Engagement Assistant (LEA), an AI tutoring system that combines retrieval-augmented generation with knowledge component models. This research extends previous work by deploying LEA in a real classroom setting with eight students and evaluating its scalability across three different courses. While LEA's core orchestration remained unchanged, the study found that synthetic evaluations alone cannot fully predict real-world performance, and faithfulness metrics declined as curriculum distance increased, suggesting further investigation into component course-agnosticism is needed. AI
IMPACT This research highlights the challenges in scaling AI tutoring systems across diverse curricula and the limitations of synthetic evaluations for predicting real-world performance.
RANK_REASON The cluster contains an academic paper detailing a new AI system and its evaluation.
- CMP511
- ICAART 2026
- Knowledge Component (KC)
- Learning Engagement Assistant (LEA)
- Ragas
- retrieval-augmented generation
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
