Learning Nonlinear Factor Models with Unknown Monotone Links from Incomplete and Noisy Data

RSS ingest of 11 new arXiv stat_ml preprints (2605.*) covering theory and methods: nonlinear factor models with monotone RKHS links and BCD recovery; ratio-based CATE estimation with doubly robust Q-Learners; identifiability guarantees for LeJEPA in latent-variable world models; a stochastic-control interpretation of CART random forests (CART-ROSA); transformers that can learn posterior predictive distributions in-context (PFNs); analysis of representational alignment governed by SNR and sample size; constrained Bayesian experimental design via amortized policies plus online planning; causal‑/