SurakshaEval / README.md
deboincepai's picture
Reduce centered summary image width to 300 pixels
6c27927 verified
|
Raw History Blame
14.1 kB
---
pretty_name: SurakshaEval
language:
- en
- as
- bn
- gu
- hi
- kn
- ml
- mr
- pa
- ta
- te
license: "cc-by-nc-sa-4.0"
tags:
- safety
- multilingual
- evaluation
- benchmark
- EN/IN scripts
- regional context
configs:
- config_name: all
default: true
data_files:
- split: test
path:
- data/assamese/test.jsonl
- data/bengali/test.jsonl
- data/gujarati/test.jsonl
- data/hindi/test.jsonl
- data/kannada/test.jsonl
- data/malayalam/test.jsonl
- data/marathi/test.jsonl
- data/punjabi/test.jsonl
- data/tamil/test.jsonl
- data/telugu/test.jsonl
- config_name: assamese
data_files:
- split: test
path:
- data/assamese/test.jsonl
- config_name: bengali
data_files:
- split: test
path:
- data/bengali/test.jsonl
- config_name: gujarati
data_files:
- split: test
path:
- data/gujarati/test.jsonl
- config_name: hindi
data_files:
- split: test
path:
- data/hindi/test.jsonl
- config_name: kannada
data_files:
- split: test
path:
- data/kannada/test.jsonl
- config_name: malayalam
data_files:
- split: test
path:
- data/malayalam/test.jsonl
- config_name: marathi
data_files:
- split: test
path:
- data/marathi/test.jsonl
- config_name: punjabi
data_files:
- split: test
path:
- data/punjabi/test.jsonl
- config_name: tamil
data_files:
- split: test
path:
- data/tamil/test.jsonl
- config_name: telugu
data_files:
- split: test
path:
- data/telugu/test.jsonl
dataset_info:
- config_name: all
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: assamese
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: bengali
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: gujarati
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: hindi
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: kannada
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: malayalam
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: marathi
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: punjabi
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: tamil
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
- config_name: telugu
features:
- name: id
dtype: string
- name: index
dtype: int64
- name: region
dtype: string
- name: language
dtype: string
- name: question_en
dtype: string
- name: question_indic
dtype: string
- name: risk_area
dtype: string
- name: type_of_harm
dtype: string
- name: question_type
dtype: string
---
# SurakshaEval
Existing LLM safety benchmarks focus primarily on English and Western contexts,
often overlooking the linguistic diversity and culturally grounded safety risks
present in other languages.
We introduce SurakshaEval, a novel safety benchmark composed of human-written
prompts spanning real-world scenarios, explicitly designed for ten major Indian
languages - Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi,
Punjabi, Tamil, and Telugu, along with English.
We benchmark state-of-the-art LLMs on SurakshaEval, establishing safety baselines
and identifying recurring failure modes, including over-refusal and insufficient
contextual awareness in regionally sensitive settings, particularly in native scripts.
These findings underscore the need for region-specific data and structured
evaluations to build safe, ethical, and culturally aligned AI systems.
This release contains **1,484 English/Indic question pairs (2,968 prompts)**
across **10 regional Indic languages**, alongside English. Each row contains a
**question pair**: an English prompt (`question_en`) and its corresponding
Indic-language prompt (`question_indic`) describing the same scenario in that
regional context. Each pair therefore contributes two prompts, one in each
language. See the
[published paper](https://ojs.aaai.org/index.php/AIES/article/view/43143) for the
benchmark methodology and evaluation results.
![Warning: This dataset contains potentially offensive or unsafe content.](diagrams/content_warning.svg)
<p align="center">
<img src="diagrams/summary_v2.png" alt="SurakshaEval summary diagram" width="300" />
</p>
## Access requirements
SurakshaEval is a **gated dataset**. Users must provide their contact information
(**email address and Hugging Face username**) and agree to the terms and
conditions, if any, presented on Hugging Face in order to access this dataset.
Sign in to Hugging Face and submit an access request on the
[dataset page](https://huggingface.co/datasets/SurakshaAI/SurakshaEval).
Once access is granted, use an access token belonging to that same account
for the loading and downloading of examples below.
## Getting started
Install the libraries for loading and downloading the dataset:
```bash
pip install datasets huggingface_hub
```
After access is granted, set that account's Hugging Face
[access token](https://huggingface.co/settings/tokens) before starting Python:
```bash
export HF_TOKEN="hf_your_access_token"
```
Replace the placeholder with your token. Hugging Face libraries automatically
use [HF_TOKEN](https://huggingface.co/docs/huggingface_hub/package_reference/environment_variables)
for authentication.
The examples below use the Hugging Face repository `SurakshaAI/SurakshaEval`.
### Load and access the dataset
```python
from datasets import load_dataset
REPO_ID = 'SurakshaAI/SurakshaEval'
# Load every region or select an Indic-language configuration.
pairs = load_dataset(REPO_ID, "all", split="test")
regional_pairs = load_dataset(REPO_ID, 'assamese', split="test")
# Access the aligned questions and annotations.
example = pairs[0]
print(example["question_en"])
print(example["question_indic"])
print(example["language"], example["type_of_harm"])
english_questions = pairs["question_en"]
indic_questions = pairs["question_indic"]
```
### Download the full dataset
Download all repository files, then load the question pairs from the JSONL files:
```python
from pathlib import Path
from datasets import load_dataset
from huggingface_hub import snapshot_download
download_dir = snapshot_download(
repo_id='SurakshaAI/SurakshaEval',
repo_type="dataset",
local_dir="surakshaeval",
)
files = sorted(str(path) for path in Path(download_dir).glob("data/*/test.jsonl"))
pairs = load_dataset("json", data_files={"test": files}, split="test")
```
### Download one regional file
For example, download and load only the Assamese-context question pairs:
```python
from datasets import load_dataset
from huggingface_hub import hf_hub_download
path = hf_hub_download(
repo_id="SurakshaAI/SurakshaEval",
repo_type="dataset",
filename="data/assamese/test.jsonl",
local_dir="surakshaeval",
)
regional_pairs = load_dataset("json", data_files={"test": path}, split="test")
```
### Save a few example pairs
Save the first five Assamese-context English/Indic pairs to a UTF-8 JSONL file:
```python
import json
from datasets import load_dataset
sample_pairs = load_dataset(
"SurakshaAI/SurakshaEval", "assamese", split="test[:5]"
)
with open("surakshaeval_example_pairs.jsonl", "w", encoding="utf-8") as file:
for pair in sample_pairs:
file.write(json.dumps(pair, ensure_ascii=False) + "\n")
```
Replace `assamese` with another regional configuration, such as `hindi`, or
use `all` to select from the combined dataset. For a single-file download,
also change the region in `filename`.
See the Hugging Face guides for
[loading datasets](https://huggingface.co/docs/datasets/loading) and
[downloading dataset files](https://huggingface.co/docs/huggingface_hub/guides/download).
## Dataset structure
The default `all` configuration includes every region. Each region also has a
configuration of its own. All configurations use a single `test` split because
this release is intended for benchmark evaluation. They reference the same
JSONL files, so the `all` configuration does not duplicate data on disk.
| Region | Indic language | Question pairs |
| --- | --- | ---: |
| assamese | as | 128 |
| bengali | bn | 149 |
| gujarati | gu | 110 |
| hindi | hi | 116 |
| kannada | kn | 163 |
| malayalam | ml | 126 |
| marathi | mr | 129 |
| punjabi | pa | 195 |
| tamil | ta | 157 |
| telugu | te | 211 |
| **Total** | | **1,484** |
| Column | Type | Meaning |
| --- | --- | --- |
| `id` | string | Stable pair identifier: Indic ISO language code and source index. |
| `index` | int64 | Original index within the region; not globally unique. |
| `region` | string | Source folder/context, such as `hindi`. |
| `language` | string | Indic ISO 639-1 language code, such as `hi`; English is always `en`. |
| `question_en` | string | Original English question, preserved verbatim. |
| `question_indic` | string | Original Indic-language question, preserved verbatim. |
| `risk_area` | string | Original risk-area annotation. |
| `type_of_harm` | string | Original harm category. |
| `question_type` | string | Original question-type annotation; blank values remain empty strings. |
## Intended use and limitations
Use the paired questions to compare safety behavior across English and the
corresponding Indic language while keeping regional context fixed. The prompts
include harmful or offensive content for evaluation of LLMs.
Keep each English/Indic pair together if creating other splits. Questions from
different regions may overlap and are retained as distinct benchmark examples.
## License
The SurakshaEval dataset is licensed under
[Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International
(CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/).
The Hugging Face license identifier is `cc-by-nc-sa-4.0`.
Sharing and adapting the dataset is permitted for noncommercial purposes
with appropriate attribution, a license link, and an indication of changes.
Shared adaptations must use the same license. See the
[full license terms](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode.en).
## Citation
If you use the SurakshaEval dataset, benchmark, or code, please cite our
[paper](https://ojs.aaai.org/index.php/AIES/article/view/43143) using the BibTeX entry below:
DOI: [10.1609/aies.v9i1.43143](https://doi.org/10.1609/aies.v9i1.43143)
```bibtex
@article{banerjee2026surakshaeval,
title = {{SurakshaEval}: {A}n {I}ndic {S}afety {B}enchmark for {M}ultilingual {LLMs}},
author = {Banerjee, Debopriyo and Kavitha, Kapil Rajesh and Borah, Angana and
Han, Xudong and Wang, Yuxia and Krishnamurthy, Parameswari and
Agarwal, Utkarsh and Kulkarni, Atharva and Lata, Swaran and
Munot, Ayush and Sahnan, Dhruv and Singh, Aaryamonvikram and
Nakov, Preslav and Choudhury, Monojit},
journal = {Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society},
volume = {9},
number = {1},
pages = {325--336},
year = {2026},
month = oct,
doi = {10.1609/aies.v9i1.43143},
url = {https://doi.org/10.1609/aies.v9i1.43143}
}
```