Axiom ASE 2026 evaluation dataset

Data, harness, and reproduction scripts for the ASE 2026 Industry Showcase paper "From Context to Code: Codifying Cross-System Reasoning for LLM-Assisted Operations". Includes a 270-cell compiled evaluation dataset (9 models × 3 passes × 5 queries × 2 modes), five evaluation queries with correctness rubrics, per-query and per-model aggregations, LLM-as-judge verdicts with a sanitized notes column, and a one-command reproduction script that regenerates the paper's figures and per-model results table. Tenant and cluster names are sanitized to placeholders. Axiom engine internals and tenant-specific data sources are not included; all quantitative claims in the paper are reproducible from the released dataset and scripts.