BlueRange: Reproducibility Data and Artifact for Authority-Aware Evaluation of Autonomous LLM Cyber Defenders

BlueRange is an open-source benchmark for evaluating the defensive effectiveness and operational safety of autonomous LLM-based cyber defenders under explicit authority constraints. This artifact contains curated experimental data, provenance information, documentation, and deterministic regeneration scripts supporting the study "BlueRange: Evaluating the Safety–Effectiveness Trade-off of Autonomous LLM Cyber Defenders." The released data support the reported results across the evaluated scenarios, model treatments, authority levels, evidence profiles, and protocol-reliability measurements. The included regeneration script reproduces the corresponding aggregate tables from the released canonical data. Provider-mediated model execution may not be byte-for-byte reproducible because of model/runtime availability and provider nondeterminism.