R18 evidence archive: result files for "A Correct Forecast Can Fail Its Own Test: Diagnosing and Repairing Finite-Simulation Spatial Consistency Tests for Earthquake Forecasts"

This archive holds the 29 result files behind the quantitative claims in "A Correct Forecast Can Fail Its Own Test: Diagnosing and Repairing Finite-Simulation Spatial Consistency Tests for Earthquake Forecasts" (Xiaofeng Shi and Hua Zhou, Beijing Academy of Artificial Intelligence). The files are the per-panel records, per-day scores, summaries and predictions under results/runs/ that the paper's 31-claim registry cites but that the code repository does not ship. The claim registry, the frozen protocols and the code are in the companion repository https://github.com/xiaofengShi/quakerank (MIT for code, CC BY 4.0 for the paper). Contents: 29 files, 90.6 MB uncompressed, 14.6 MB zipped, SHA-256 50445a0641891dae102b2da048f88df8cb2827113ed5020a394586f5efb394ac. Paths are repository-relative. Extract over the root of the companion repository and verify: unzip evidence_archive_20261104_v1.zip -d /path/to/quakerankcd /path/to/quakerankpython scripts/build_evidence_manifest.py --verify The check reports 43/43 files matching evidence_manifest.json. SHA256SUMS lists every file in the archive. No SMASH code or weights are included; they are not redistributable. The Italian catalog snapshot (INGV, CC BY 4.0) ships with the repository, not this archive; ComCat is U.S. public domain.