Assess whether regime composition explains prior domain-level differences

Determine whether the domain-level differences previously reported for aggregate and LLM-generated user profiles are explained by differences in the composition of habitual and exploratory consumption regimes.

Background

The current paper finds that profile effectiveness depends on the user’s consumption regime: aggregate profiles perform better for habitual users, whereas LLM-generated profiles perform better for exploratory users. Earlier evaluations reported differences between content domains but did not distinguish users according to these regimes.

Consequently, it remains unresolved whether those earlier domain-level performance differences reflect intrinsic domain effects or simply different proportions of habitual and exploratory users in the evaluated datasets.

References

The earlier evaluations on public data did not separate the two regimes, so whether the domain-level differences reported there reduce to differences in regime composition remains untested.

— When LLM-Inferred User Context Adds Value in Production Streaming Recommendation  (2609.38999 - Sabouri et al., 30 Sep 2026) in Section 5.3, “Limitations”