Skip to content

Dataset · facebook

KILT benchmark on Wikipedia

facebook/kilt_tasks

11 datasets for fact-checking, QA, entity linking, and dialogue on a single Wikipedia dump.

What's inside

KILT combines 11 datasets across five task types: fact-checking, entity linking, slot-filling, open-domain QA, and dialogue generation. All are tied to a single preprocessed English Wikipedia dump. This provides uniform evaluation and enables multitask and transfer learning.

What you can do with it

  • Train a multitask model for QA and retrieval
  • Evaluate RAG systems on KILT tasks
  • Test entity linking and fact-checking
Tasks
Fill mask, Question answering, Text classification, Text generation, Text retrieval
Languages
English
Size
1M–10M rows
Downloads this month
7.1K
Likes
69

How to load it

from datasets import load_dataset

ds = load_dataset("facebook/kilt_tasks")