metadata
license: mit
task_categories:
- table-question-answering
pretty_name: Hybrid RAG Architecture Evaluation Corpus
size_categories:
- n<1K
This dataset comprises the five core reference documents utilized for the qualitative evaluation of a Hybrid Retrieval-Augmented Generation (RAG) Architecture. To rigorously assess the system's performance across diverse document structures—including text, tables, and metadata—these files were selected as the primary knowledge base for testing.
Evaluation was conducted using expert human judgment to ensure the architectural outputs were factually grounded, contextually relevant, and free from hallucinations. This curated corpus serves as a standardized baseline for benchmarking RAG systems that require high-fidelity retrieval from complex, multi-modal PDF sources