Probabilistic Residual Learning for Online Recommendations

Collection of recent research (Jul 2026) on recommender systems, retrieval, and LLM-augmented pipelines. Papers cover probabilistic residual learning (PRL) for causal refinement of recommenders, cost-aware sparse feature ranking (LO-FAR), controllable content-based recommenders (CCBR), ANNS optimizations for high-dimensional LLM embeddings, generative recommendation architectures (BARGE, DLMRec), LLM-based retrievers and training frameworks (SHIFT), LLM+knowledge-graph navigation (Search-on-Graph, Search-on-Graph-R1), an evidence-aware multi-skill agent for literature analysis (AlphaAgent),and