Prediction Is Not Calibration: Item Parameter Recovery and Representation Coupling in Knowledge Tracing

Not provided

knowledge tracingitem response theoryparameter recovery

2610.30211

Builder Relevance

70%

2h ago

Abstract

This paper investigates the limitations of knowledge tracing models in accurately recovering item parameters despite high predictive accuracy.

Reality Card

Core Claim

The study demonstrates that item parameters, particularly discrimination and option slopes, must be evaluated separately from response prediction to improve recovery and consistency.

Method / Result

Widening the item parameter key ($d_M$) consistently improves agreement with offline IRT reference across various datasets.

Limitations

The trade-off from narrowing the sequence item representation ($d_S$) depends on the encoder's reliance on sequence width for prediction, which may affect reproducibility.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.

← Back to all papers