Joy Bose, Om Thomas (2026) — arXiv preprint 📄 Full text (arXiv)
Standardized examinations are typically treated as uniform syllabus coverage problems. LearnOpt recovers stable latent cognitive structures diverging systematically from official syllabi, using LLM-tagged questions and constrained optimization. Applied to 9 years of NEET questions (n=1,496) and JEE Advanced questions. Finds NEET latent skill distribution is stable within syllabus regimes (KL 0.004-0.032) but shifts significantly after syllabus rationalization (KL=0.040, p=0.0005). JEE Advanced is dominated by multi-concept integration (80.9%). Formulates study planning as a knapsack-variant optimization with Bayesian Knowledge Tracing.
Key Contributions
- Standardized exams have recoverable latent cognitive structures; multi-concept integration dominates higher-tier exams (JEE: 80.9%).
Related Pages
- ai-k12-evidence-base — Empirical evidence on AI in K-12 education outcomes
- intelligent-tutoring — Automated tutoring systems and their evaluation
- student-experience — Student perspectives on AI in education
- learning-analytics — Data-driven approaches to understanding learning
- llm-feedback-programming-classroom — LLM feedback in classroom settings
- student-cheat-sheets-make-or-take — Students choose between self-created and instructor-provided cheat sheets based on trust, personaliz
Citation
APA: Joy Bose, Om Thomas (2026). LearnOpt: Recovering the Latent Cognitive Structure of Standardized Examinations via Knowledge Graphs and Constrained Optimization. arXiv:2606.15349. arXiv preprint.