Anonymous

This record contains an anonymized replication package for a study on prompt-based LLM approaches for software requirements classification, covering: RQ1.1: 4-way requirement-type overlap classification (Functional / Non-Functional / Both / Neither) RQ1.2: 12-category ISO-aligned multi-label NFR subtype prediction RQ2: Rationale quality evaluation using an LLM-as-judge pairwise tournament protocol RQ3: Rationale-augmented prompting comparison (static vs dynamic rationale selection via semantic similarity) Package structure RQ1.1/ — outputs and evaluation summaries for the 4-way task RQ1.2/ — outputs and evaluation summaries for the multi-label subtype task RQ2/ — tournament match outputs and selected-rationale results RQ3/ — results for static vs dynamic rationale augmentation prompts.pdf — prompt templates used across all experiments