Dataset for Agentic Knowledge Curation versus Full-Context Retrieval: an empirical study of retrieval failure topology in long-context LLM Systems

This repository contains the supplementary materials for a study that compares two knowledge management paradigms — Full-Context Retrieval and LLM-Wiki — over the technical documentation of RJusticia, a semantic search system for Spanish legal documents. Evaluation was performed by two independent external reviewers in a blind setup. The files included contain the evaluation corpus, the full question set with ground-truth answers, the prompts used in both experimental conditions, the generated answers from both systems, the blind reviewer scores, and the curated knowledge base built by the LLM-Wiki agent.