Datasets:
Verification data guide
data/verification_cases.jsonl contains 157 records derived from the four unmodified reports in results/recorded/.
| Field | Type | Meaning |
|---|---|---|
case_id |
string | Stable suite name and one-based case index |
suite |
string | rigidity, real_minimax, extensions, or complex |
kind |
string | Original case category |
case_name |
string | Original name/support, or category when unnamed |
passed |
boolean | Inherits the source report's all_checks_passed status |
tolerance |
number | Source suite's recorded numerical tolerance |
source_report |
string | Repository path to the full source report |
details_json |
string | Original complete case object serialized as JSON |
passed reports recorded execution status. It does not establish external replication, theorem validity, or worldwide novelty. Exact combinatorial examples and floating-point computations appear together; kind and details_json distinguish their recorded methods.
data/verification_summary.csv contains one row per suite, including case count, tolerance, random seed, and report path. Solver statuses and maximum residuals remain in the original JSON reports.
The Hub configuration explicitly selects the JSONL file, preventing heterogeneous scientific report schemas from being inferred as one table. Reproduction writes fresh results separately to results/regenerated/.