DAPFAM: Patent Family Retrieval Benchmark Tasks

DAPFAM is a dataset for patent family retrieval tasks, now integrated into the MTEB benchmark. The dataset page was last updated on September 10, 2025, by the author 'datalyes'. It includes 18 distinct retrieval tasks derived from different query and target views.

Use Cases

• Benchmarking embedding models on patent retrieval tasks based on the 18 integrated MTEB tasks.
• Reproducing research paper results based on the specified Snowflake Arctic Embed v2 encoder.
• Evaluating model performance on different query and target views of patent family data.

Strengths

• Integrated into the MTEB benchmark, enabling standardized evaluation.
• Includes 18 distinct retrieval tasks for comprehensive benchmarking.
• Paper results are reproducible using the specified Snowflake/snowflake-arctic-embed-m-v2.0 encoder.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

datalyes

Freshness

Last updated 2025-09-10 11:55:52.