RegimeMamba: Regime-Sensitive Conditional Diffusion for Multi-Asset Probabilistic Forecasting

Anonymous reproducibility artifact for RegimeMamba, an end-to-end framework for probabilistic multi-asset return forecasting. The archive includes the paper-compatible selective state-space implementation, differentiable discrete conditioning module, conditional diffusion decoder, training and evaluation code, controlled architecture ablations, matched Transformer and score-SDE baselines, traditional and univariate controls, conformal calibration procedures, cross-market evaluation scripts, configurations, tests, and lightweight CSV/JSON experimental evidence. Licensed market data and large model checkpoints are not redistributed; data reconstruction and checkpoint requirements are documented in the archive. All author-identifying information, local paths, submission metadata, and reviewer materials have been removed.