Airavata Evaluation Suite
Collection
A collection of benchmarks used for evaluation of Airavata, an Hindi instruction-tuned model on top of Sarvam's OpenHathi base model.
•
22 items
•
Updated
•
6
Datasets:
Tasks:
Question Answering
Sub-tasks:
closed-domain-qa
Multilinguality:
multilingual
Size Categories:
n<1K
Language Creators:
found
Annotations Creators:
expert-generated
Source Datasets:
original
License:
Dataset Viewer
View in Dataset Viewer
Viewer
The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider
removing the
loading script
and relying on
automated data support
(you can use
convert_to_parquet
from the datasets library). If this is not possible, please
open a discussion
for direct help.
Dataset Card for [Dataset Name]
Description: The Indic QA dataset is designed for question answering tasks, with a focus on Indic languages. It contains questions paired with corresponding contexts and answers. The dataset aims to facilitate research and development in question answering systems for Indic languages.
Dataset Info:
Features:
- ID: Identifier for each data instance.
- Context: The passage or context providing information relevant to answering the question.
- Question: The question posed by the user.
- Answers: The possible answers to the question, provided as a sequence.
Languages:
- Assamese (as)
- Bengali (bn)
- Hindi (hi)
- Kannada (kn)
- Marathi (mr)
- Malayalam (ml)
- Punjabi (pa)
- Oriya (or)
- Tamil (ta)
- Telugu (te)
Thanks to @github-username for adding this dataset.
- Downloads last month
- 2,001