Uniform teacher-quality effects across student backbones

Determine whether the teacher-quality effect observed for gpt-5-mini and Qwen2.5 3B holds uniformly across all ten student backbones evaluated in the rEDMRec recommendation experiments.

Background

The study examines the effect of teacher quality by varying the teacher while fixing the student to either gpt-5-mini or Qwen2.5 3B. These experiments indicate that teacher-generated memory quality, particularly lower duplication, can improve downstream ranking, but the evaluation does not cover every teacher–student pairing.

Consequently, it remains unresolved whether the observed relationship generalizes consistently to all ten student models used in the main experiments. Establishing this would clarify the robustness of rEDMRec’s teacher-quality findings across student capacities and architectures.

References

The main results (Section~\ref{sec:results-h1}) fix the teacher to gpt-5.4-mini; the teacher-distillation study (Section~\ref{sec:results-h3}) varies the teacher but only against two fixed students. We have not measured the full teacher $\times$ student cross-product, so it remains open whether the teacher-quality effect observed for gpt-5-mini and Qwen2.5 3B holds uniformly across all ten students in Table~\ref{tbl:app-full}.

— rEDMRec: Distilling Large Language Model Reasoning into an Editable Experience Memory for Recommendation  (2608.18952 - Nguyen et al., 19 Aug 2026) in Discussion and Limitations, subsection “Teacher coverage”