Knowledge Tracing
A core task in educational technology that models students' historical learning behaviors to track and predict dynamic changes in their knowledge mastery states. It forms the foundational technology for adaptive learning systems and intelligent tutoring systems.
Timeline (last 90 days)
A study evaluated six knowledge tracing models on four public educational datasets and tested 12 threshold levels ranging from 0.50 to 0.99.
The study searched for balanced thresholds across 30 preset instructional scenarios, with results showing that the optimal balanced threshold varies greatly across different models and scenarios.
Stricter thresholds do not reliably narrow performance gaps between students, and may instead disproportionately restrict student advancement.
In certain settings, students with stronger prior ability advance at a rate up to 3.26 times higher than students with weaker prior ability.
The quality of a threshold should be evaluated comprehensively across four dimensions—performance, practice, advancement, and fairness—rather than by a single metric alone.
A mastery threshold is not a magic number that can be reused across models.
All Facts (6)
A study evaluated six knowledge tracing models on four public educational datasets and tested 12 threshold levels ranging from 0.50 to 0.99.
50%UnverifiedThe study searched for balanced thresholds across 30 preset instructional scenarios, with results showing that the optimal balanced threshold varies greatly across different models and scenarios.
50%UnverifiedStricter thresholds do not reliably narrow performance gaps between students, and may instead disproportionately restrict student advancement.
50%UnverifiedIn certain settings, students with stronger prior ability advance at a rate up to 3.26 times higher than students with weaker prior ability.
50%UnverifiedThe quality of a threshold should be evaluated comprehensively across four dimensions—performance, practice, advancement, and fairness—rather than by a single metric alone.
50%UnverifiedA mastery threshold is not a magic number that can be reused across models.
50%