A Quantum-Classical Hybrid Framework for Multivariate Time-Series Forecasting Complexity-Fidelity Trade-offs and Limitations

Collection of recent ML/IR papers (arXiv 2026-07-21) focused on recommender systems, retrieval, multimodal cold-start, uncertainty-aware ranking, scalable architectures, and a quantum-classical hybrid framework for multivariate time-series forecasting. Key contributions include: WHALE (unified Wukong+HSTU recommendation backbone; deployed in production), BONSAI (decoding-trie optimization for LLM-based generative recommendation), UAME (uncertainty-aware ensemble ranking to mitigate label bias; deployed), sparse multimodal embeddings for cold-item recommendation, Matryoshka Hypencoder (multi‑s‑