Knowledge tracing (KT) models students’ knowledge states and predicts their future performance based on their historical interaction data. Our research identifies two significant limitations in current attention based KT approaches: they inadequately differentiate between content relevance and memory decay patterns, and their fixed-context windows cannot adapt to increasingly lengthy learning sequences. This paper presents LFPKT, a novel framework that enhances attention based KT by implementing adaptive forgetting mechanisms. By introducing position-sensitive attention calculations, LFPKT separates knowledge relevance from temporal forgetting effects. This separation allows the model to accurately capture various forgetting behaviors while maintaining performance across sequences of different lengths.