Effect of training across many recommendation scenarios

Determine whether training the Decision4Rec decision model on many more recommendation scenarios would improve its performance.

Background

Decision4Rec trains the Kev decision model on recommendation records from three source domains and evaluates transfer to two held-out domains. The paper reports results for several model sizes and notes that adding a fourth training domain did not improve performance. However, the experiments do not establish whether substantially increasing the number of training scenarios would benefit the decision model, leaving the scalability of cross-scenario training unresolved.

References

Training on a fourth domain did not help . Whether many more scenarios would help is not tested.

— Decision4Rec: A Jev-Inspired Framework of Decision Models on Knowledge and Active Graphs for Multi-Agent Recommender Systems  (2610.26035 - Polignano et al., 19 Oct 2026) in Section Discussion, paragraph “Few domains, few sizes, one run”