A reproducible and extensible framework for benchmarking competing risks survival models

The document is an arXiv statistics and machine-learning feed containing research on survival-model benchmarking, causal inference with unstructured treatments, diffusion-model rollout error detection, grouped-feature FDR control, distribution-shift detection, bandit identification, finite-basis distribution certificates, temporal learning limits, likelihood-equation nonproperness, and private generative bootstrap methods. It contains no apparent cybersecurity incident, exploit, malware, vulnerability disclosure, or directly actionable threat intelligence.