---
license: apache-2.0
task_categories:
- question-answering
language:
- en
size_categories:
- n<1K
---
In this collection you can find 4 datasets with `is_supporting=True` contexts from the Adaptive RAG collection. 
There are picked 4/6 datasets from Adaptive RAG datasets with `is_supporting=True` contexts. 
Not all samples from TriviaQA and SQUAD have `is_supporting=True` contexts, thats why we do not include them in hf collection.
If question have more than one `is_supporting=True` context, we concatenate them. 
Script for data transformation from original Adaptive RAG format into our format can be found here:
https://github.com/sashaboriskin/rag_routing/blob/main/data/make_hf_adaptive_rag_datasets.py