Bayesian Latent Space Models for Graphs Are Misspecified: Toward Robust Inference via Generalized Posteriors

Collection of recent ML/theory papers (arXiv 20 May 2026) covering robust inference for graph latent-space models (showing misspecification causes overconfidence and proposing a generalized posterior — Link-Sequential R-SafeBayes), heavy-tail generation limits of Gaussian/Lipschitz decoders and a Phase-Type (Markov chain) decoder solution, a transported-Beta-law view of split conformal prediction with finite-sample bounds, data‑driven Lagrangian relaxation for MILP with minimax/generalization rates and SGA optimality, a Dual-Channel Tensor Neural Network (low-rank + sparse refinement) with non