Replication Data for: Knowing When to Escalate: Prompt-Based Uncertainty Quantification for Selective Automation of LLM-Based Intrusion Detection

Replication package for the Array article "Knowing When to Escalate: Prompt-Based Uncertainty Quantification for Selective Automation of LLM-Based Intrusion Detection" (doi:10.1016/j.array.2026.101127). Version 2.1.0 contains the full Phase 2 evaluation: raw per-sample outputs of 70 LLM-based intrusion detection experiments (7 prompt-based uncertainty quantification methods x 5 LLMs x 2 CIC-IDS benchmark datasets, 750 stratified samples each), the baseline attack-blind label-leakage ablation, the evaluation datasets, the experiment code, and pre-computed aggregated metrics. Version 1.0.0 contained the 50-sample Phase 0 pilot only. Version 2.1.0 adds CORRECTION.md, which corrects the article (the stochastic methods' correctness was scored against the attack type as well as the binary label; the label leak affects CIC-IDS2017 only; the provider of some Kimi K2 cells), and recompute/, scripts that recompute the corrected metrics from the unchanged raw outputs.This version (October 2026). All files of version 2.1.0 (DOI 10.5281/zenodo.23586200) are preserved unchanged. The added folder supervisor-2026-10/ holds the analyses, inputs, hashes and limitations from the supervisor review of the PhD thesis; read its README first. It reports no new model evaluation.