Pointwise is Pointless? A Multimodal Ablation Study for Precipitation Nowcasting with Graph Neural Networks

This feed (arXiv 18 Jun 2026) collects nine stat/ML papers. Key items: (1) A multimodal graph-neural-network nowcasting study showing how sparse point observations (Netatmo), NWP (MEPS), and satellite inputs differentially improve short-term precipitation forecasts; (2) ToolChain-CRC — a conformal risk-control framework for retrieval-augmented and tool-using agentic systems that scores trajectories, calibrates accept/intervene rules, and provides an anytime alarm with provable trajectory-level risk guarantees and drift-aware extensions; (3) Compact geometric reachability embeddings for hierarc