---
license: apache-2.0
dataset_info:
- config_name: arc_challenge
  features:
  - name: example_id
    dtype: string
  - name: query
    dtype: string
  - name: choices
    list: string
  - name: gold
    dtype: int64
  splits:
  - name: test
    num_bytes: 362046
    num_examples: 1172
  download_size: 275418
  dataset_size: 362046
- config_name: arc_easy
  features:
  - name: example_id
    dtype: string
  - name: query
    dtype: string
  - name: choices
    list: string
  - name: gold
    dtype: int64
  splits:
  - name: test
    num_bytes: 628105
    num_examples: 2376
  download_size: 463671
  dataset_size: 628105
- config_name: boolq
  features:
  - name: example_id
    dtype: string
  - name: question
    dtype: string
  - name: passage
    dtype: string
  - name: label
    dtype: int64
  splits:
  - name: test
    num_bytes: 2272326
    num_examples: 3270
  download_size: 1471811
  dataset_size: 2272326
- config_name: hellaswag
  features:
  - name: example_id
    dtype: string
  - name: query
    dtype: string
  - name: choices
    list: string
  - name: gold
    dtype: int64
  splits:
  - name: test
    num_bytes: 8540674
    num_examples: 10042
  download_size: 7301658
  dataset_size: 8540674
- config_name: nq_open
  features:
  - name: example_id
    dtype: string
  - name: question
    dtype: string
  - name: answer
    list: string
  splits:
  - name: test
    num_bytes: 404421
    num_examples: 3610
  - name: train
    num_bytes: 9091871
    num_examples: 87925
  download_size: 5917498
  dataset_size: 9496292
- config_name: openbookqa
  features:
  - name: example_id
    dtype: string
  - name: query
    dtype: string
  - name: choices
    list: string
  - name: gold
    dtype: int64
  splits:
  - name: test
    num_bytes: 84699
    num_examples: 500
  download_size: 71491
  dataset_size: 84699
- config_name: piqa
  features:
  - name: example_id
    dtype: string
  - name: goal
    dtype: string
  - name: choices
    list: string
  - name: gold
    dtype: int64
  splits:
  - name: test
    num_bytes: 543708
    num_examples: 1838
  download_size: 478588
  dataset_size: 543708
- config_name: triviaqa
  features:
  - name: example_id
    dtype: string
  - name: question
    dtype: string
  - name: answer_value
    dtype: string
  - name: answer_aliases
    list: string
  splits:
  - name: test
    num_bytes: 7812049
    num_examples: 17944
  - name: train
    num_bytes: 58896326
    num_examples: 138384
  download_size: 58142652
  dataset_size: 66708375
- config_name: winogrande
  features:
  - name: example_id
    dtype: string
  - name: sentence
    dtype: string
  - name: option1
    dtype: string
  - name: option2
    dtype: string
  - name: answer
    dtype: string
  splits:
  - name: test
    num_bytes: 198936
    num_examples: 1267
  download_size: 102151
  dataset_size: 198936
configs:
- config_name: arc_challenge
  data_files:
  - split: test
    path: arc_challenge/test-*
- config_name: arc_easy
  data_files:
  - split: test
    path: arc_easy/test-*
- config_name: boolq
  data_files:
  - split: test
    path: boolq/test-*
- config_name: hellaswag
  data_files:
  - split: test
    path: hellaswag/test-*
- config_name: nq_open
  data_files:
  - split: test
    path: nq_open/test-*
  - split: train
    path: nq_open/train-*
- config_name: openbookqa
  data_files:
  - split: test
    path: openbookqa/test-*
- config_name: piqa
  data_files:
  - split: test
    path: piqa/test-*
- config_name: triviaqa
  data_files:
  - split: test
    path: triviaqa/test-*
  - split: train
    path: triviaqa/train-*
- config_name: winogrande
  data_files:
  - split: test
    path: winogrande/test-*
---

# Serbian LLM Eval

This dataset is a republish of
[gordicaleksa/serbian-llm-eval-v1](https://huggingface.co/datasets/gordicaleksa/serbian-llm-eval-v1)
as plain Parquet configs, so it can be loaded with modern versions of the
`datasets` library (the upstream repo ships a Python loading script that
`datasets` can no longer execute). Row contents are otherwise unchanged; only
the `example_id` field is synthesized where the upstream data has no unique
identifier, and the `triviaqa` `answer` struct is flattened into
`answer_value` / `answer_aliases` columns.

## Attribution

All content originates from
[Aleksa Gordić](https://www.linkedin.com/in/aleksagordic)'s
**Serbian LLM Eval** project: English benchmark data machine-translated to
Serbian (Google Translate), refined via GPT-4, with manual review. See the
[upstream dataset card](https://huggingface.co/datasets/gordicaleksa/serbian-llm-eval-v1)
and the
[project report](https://wandb.ai/gordicaleksa/serbian_llm_eval/reports/First-Serbian-LLM-eval---Vmlldzo2MjgwMDA5)
for full methodology and credits.

- **Upstream repo:** [gordicaleksa/serbian-llm-eval-v1](https://huggingface.co/datasets/gordicaleksa/serbian-llm-eval-v1)
- **Upstream revision:** `d9d7e1b6ff163fc659e652ebabb4dbf262dae387`
- **License:** Apache 2.0 (matches upstream)

## Splits and counts

| Config | Split | Rows |
|--------|-------|------|
| `arc_challenge` | `test` | 1172 |
| `arc_easy` | `test` | 2376 |
| `boolq` | `test` | 3270 |
| `hellaswag` | `test` | 10042 |
| `nq_open` | `test` | 3610 |
| `nq_open` | `train` | 87925 |
| `openbookqa` | `test` | 500 |
| `piqa` | `test` | 1838 |
| `triviaqa` | `test` | 17944 |
| `triviaqa` | `train` | 138384 |
| `winogrande` | `test` | 1267 |

Only `nq_open` and `triviaqa` have a `train` split; every other config has a
single `test` split.

## Usage

```python
from datasets import load_dataset

ds = load_dataset("permitt/serbian-llm-eval", "arc_challenge", split="test")
```

## Citation

Please cite the upstream dataset:

```
@article{serbian-llm-eval,
  author    = "Gordić Aleksa",
  title     = "Serbian LLM Eval",
  year      = "2023"
  howpublished = {\url{https://huggingface.co/datasets/gordicaleksa/serbian-llm-eval-v1}},
}
```

## License

Apache 2.0, matching the upstream `gordicaleksa/serbian-llm-eval-v1` dataset.