Pseudo Label NCF for Sparse OHC Recommendation: Dual Representation Learning and the Separability Accuracy Trade off

Collection of recent IR/recSys papers (Mar 2026) covering methods and systems: Pseudo Label NCF improves cold-start recommendations for online health communities via survey-derived pseudo labels (privacy-sensitive health data); gDMR/gSTM automate support-group formation using text, demographics and network embeddings (scalability and privacy implications); DIET introduces streaming dataset distillation for recommender systems to compress training data (efficiency gains but potential privacy/membership-leakage risks); an LLM-driven multimodal reranking framework addresses long-tail short-video/