CMB uses counterfactual perturbations on item embeddings, optimized by a collection of epsilon-greedy bandits, to improve recommendation diversity across differentiable and non-differentiable metrics while explaining which item features matter.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Counterfactual Multi-player Bandits for Explainable Recommendation Diversification
CMB uses counterfactual perturbations on item embeddings, optimized by a collection of epsilon-greedy bandits, to improve recommendation diversity across differentiable and non-differentiable metrics while explaining which item features matter.