HF dataset

hotchpotch / JFWIR

JFWIR is a large‑scale Japanese information‑retrieval dataset of over 64 million document–query pairs—generated from high‑quality educational web content, featuring seven query types and hard negatives—to improve contrastive learning and benchmark performance on tasks such as JQaRA, MIRACL(ja), jsquad, and JaCWIR.

Description in other languages

Statistics

Statistics as of Nov 16, 2026

Hugging Face downloads (last 30 days)
391
Hugging Face likes
4
Listed
Jun 24, 2025

Downloads (30-day window) over time

05001kNovSepJulMayMarJan 2026391
Show as a table
MonthDL (30d)
Dec 2025320
Jan 2026205
Feb 2026—
Mar 2026230
Apr 2026708
May 2026184
Jun 2026—
Jul 2026745
Aug 2026200
Sep 2026551
Oct 2026998
Nov 2026391

Downloads over the 30 days before the last snapshot of each month.

Popular in the same category

Resources in similar research fields

Learn from slides & articles