This dataset contains evaluation queries and ground-truth relevance labels used to benchmark Intelligent Search at scale. It is intended for retrieval evaluation, not model training.
Curated JSONL files with user-style queries and top-k labeled relevant messages.
Each dataset file corresponds to a distinct thematic domain (e.g., support, engineering, product, general chat).
Each row contains:
id: integer query… See the full description on the dataset page:
https://huggingface.co/datasets/dnouv/intelligent-search-benchmark.
