Analysis data — Explainable warfarin maintenance dosing (fixed-duration)

Analysis data bundle — Explainable warfarin maintenance dosing (fixed-duration) Curated analysis artifacts (figures + the CSVs every reported number traces to) for the paper "Distillable by Design: Action Forging and Reward Shaping for Interpretable Reinforcement-Learning Warfarin Dosing." This bundle is the data companion to the code snapshot. It contains no raw model checkpoints and no raw trajectory dumps — those are regenerable from the code and are excluded to keep the archive small and meaningful. Code: GitHub `anzabizadeh/warfarin_dosing` and `anzabizadeh/ReiL`, tag `explainable-warfarin-fixed-duration-v1` (research-code SHAs `9b74474` / `9c28984f`). Reproduction: see `REPRODUCE.md` in the `warfarin_dosing` repository — every file here is regenerable via the scripts documented there. All results are η = 1.0 (`reward_name: sq_dist`), 5K/1K regime, unless a file name says otherwise. Cohorts Hamberg-2007 (in-distribution) and Hamberg-2010 (OOD) are reported separately. Baselines are PGPGA (Aurora) and PGPGI (Intermountain).