MemFactory: Evaluation Data Derived from HotpotQA

A lightweight derivative release of data used to evaluate the MemFactory system. The repository contains four JSON files derived from the upstream HotpotQA dataset. The dataset was authored by nworats and last updated on March 30, 2026.

Use Cases

• Benchmarking question-answering models based on the derived HotpotQA evaluation files.
• Evaluating the effectiveness of memory mechanisms in language models based on the dataset's stated purpose.
• Testing model performance on multi-hop reasoning tasks using the converted HotpotQA subset.

Strengths

• Contains four distinct evaluation files for varied testing scenarios.
• Directly derived from the established HotpotQA dataset, suggesting a known benchmark foundation.

Limitations

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

Provenance

Source

Derived from the BytedTsinghua-SIA/hotpotqa dataset.

Collection Method

Reused and adapted from the upstream HotpotQA release.

Freshness

Last updated 2026-03-30 07:42:15; freshness should be verified.