An Information-Theoretic Evaluation Framework for Benchmark and Model Diagnosis in Knowledge Tracing
Accepted as poster at NeurIPS 2026
This work formalizes benchmark and model diagnosis through redundancy, synergy, unique information, response uncertainty, and information gain.
Recommended citation: Jiang, H.*, Wang, Z.*, Cheng, T., Li, X., Hou, M., Zheng, J., Luo, R., Guo, T., & Liu, Z. (2026). "An Information-Theoretic Evaluation Framework for Benchmark and Model Diagnosis in Knowledge Tracing." NeurIPS 2026. Accepted as poster.