Distinguishing retained knowledge from engagement without retention

Determine whether engagement with online content becomes retained, interest-driven knowledge that improves performance on knowledge items grounded in an individual participant’s own individual text corpus, rather than merely producing generic benefits from exposure to web text.

Background

The adapter increased generic knowledge-related performance but did not reproduce participants’ specific answer behavior, including their errors. The authors therefore consider two competing explanations: participants may retain knowledge related to content they engaged with, especially content associated with their interests, or the adapter may benefit from web-derived information without that information having been retained by the individual. The paper states that these explanations cannot be separated with the generalized knowledge test and proposes repeated testing with new items generated from each participant’s own individual text corpus as a possible way to discriminate them.

References

This reading remains tentative, since we cannot separate it from a generic route in which some knowledge items benefit from web text as such, independently of individual engagement. A repeated measurement with new knowledge items generated from each participant's own ITC could discriminate the two accounts: retained, interest-driven knowledge predicts that participants outperform on items grounded in their own ITC, particularly where item domain and stated interests match, whereas engagement without retention predicts that they do not.

— From Retrieval to Weights: Parametric Individualization of Small Language Models with Individual Text Corpora  (2609.10155 - Wigbels et al., 9 Sep 2026) in Section Discussion and Outlook