Search papers, labs, and topics across Lattice.
5
1
4
3
A simple averaging of multiple models can dramatically enhance sentence prediction accuracy, revealing that the crowd's insights are embedded in document-level structure rather than individual model performance.
A language model can significantly outperform naive truncation in retaining reader-relevant content, achieving a 38.4% retention rate of crowd-marked sentences.
Readers' highlighting preferences are surprisingly stable over time, with personal profiles maintaining predictive power for future selections even after nearly two years.
Readers form distinct highlighting factions within documents, but whether these factions persist across different texts remains an open question.
A logistic ranker trained on reader highlights outperforms baseline predictions, revealing that learning from real engagement data significantly enhances cold-start salience forecasting.
