README.md

---
configs:
- config_name: gsm8k_araeng
  data_files:
  - split: test
    path:
    - "gsm8k/gsm8k_araeng.csv"
- config_name: gsm8k_chieng
  data_files:
  - split: test
    path:
    - "gsm8k/gsm8k_chieng.csv"
- config_name: gsm8k_hineng
  data_files:
  - split: test
    path:
    - "gsm8k/gsm8k_hineng.csv"
- config_name: gsm8k_spaeng
  data_files:
  - split: test
    path:
    - "gsm8k/gsm8k_spaeng.csv"
- config_name: lid_chieng
  data_files:
  - split: test
    path:
    - "lid/lid_chieng.csv"
- config_name: lid_fridut
  data_files:
  - split: test
    path:
    - "lid/lid_fridut.csv"
- config_name: lid_gereng
  data_files:
  - split: test
    path:
    - "lid/lid_gereng.csv"
- config_name: lid_guaspa
  data_files:
  - split: test
    path:
    - "lid/lid_guaspa.csv"
- config_name: lid_hineng
  data_files:
  - split: test
    path:
    - "lid/lid_hineng.csv"
- config_name: lid_hokman
  data_files:
  - split: test
    path:
    - "lid/lid_hokman.csv"
- config_name: lid_mareng
  data_files:
  - split: test
    path:
    - "lid/lid_mareng.csv"
- config_name: lid_msaea
  data_files:
  - split: test
    path:
    - "lid/lid_msaea.csv"
- config_name: lid_nepeng
  data_files:
  - split: test
    path:
    - "lid/lid_nepeng.csv"
- config_name: mmlu_araeng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_araeng.csv"
- config_name: mmlu_beneng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_beneng.csv"
- config_name: mmlu_chieng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_chieng.csv"
- config_name: mmlu_duteng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_duteng.csv"
- config_name: mmlu_freeng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_freeng.csv"
- config_name: mmlu_gereng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_gereng.csv"
- config_name: mmlu_hineng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_hineng.csv"
- config_name: mmlu_mareng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_mareng.csv"
- config_name: mmlu_nepeng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_nepeng.csv"
- config_name: mmlu_spaeng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_spaeng.csv"
- config_name: mmlu_tameng
  data_files:
  - split: test
    path:
    - "mmlu/mmlu_tameng.csv"
- config_name: mt_araeng_eng
  data_files:
  - split: test
    path:
    - "mt/mt_araeng_eng.csv"
- config_name: mt_beneng_eng
  data_files:
  - split: test
    path:
    - "mt/mt_beneng_eng.csv"
- config_name: mt_chieng_chi
  data_files:
  - split: test
    path:
    - "mt/mt_chieng_chi.csv"
- config_name: mt_chieng_eng
  data_files:
  - split: test
    path:
    - "mt/mt_chieng_eng.csv"
- config_name: mt_hineng_eng
  data_files:
  - split: test
    path:
    - "mt/mt_hineng_eng.csv"
- config_name: mt_hokman_man
  data_files:
  - split: test
    path:
    - "mt/mt_hokman_man.csv"
- config_name: mt_mareng_eng
  data_files:
  - split: test
    path:
    - "mt/mt_mareng_eng.csv"
- config_name: mt_spaeng_eng
  data_files:
  - split: test
    path:
    - "mt/mt_spaeng_eng.csv"
- config_name: ner_guaspa
  data_files:
  - split: test
    path:
    - "ner/ner_guaspa.csv"
- config_name: ner_hineng
  data_files:
  - split: test
    path:
    - "ner/ner_hineng.csv"
- config_name: ner_msaea
  data_files:
  - split: test
    path:
    - "ner/ner_msaea.csv"
- config_name: ner_spaeng
  data_files:
  - split: test
    path:
    - "ner/ner_spaeng.csv"
- config_name: pos_chieng
  data_files:
  - split: test
    path:
    - "pos/pos_chieng.csv"
- config_name: pos_fridut
  data_files:
  - split: test
    path:
    - "pos/pos_fridut.csv"
- config_name: pos_hineng
  data_files:
  - split: test
    path:
    - "pos/pos_hineng.csv"
- config_name: pos_spaeng
  data_files:
  - split: test
    path:
    - "pos/pos_spaeng.csv"
- config_name: sa_beneng
  data_files:
  - split: test
    path:
    - "sa/sa_beneng.csv"
- config_name: sa_hineng
  data_files:
  - split: test
    path:
    - "sa/sa_hineng.csv"
- config_name: sa_maleng
  data_files:
  - split: test
    path:
    - "sa/sa_maleng.csv"
- config_name: sa_mareng
  data_files:
  - split: test
    path:
    - "sa/sa_mareng.csv"
- config_name: sa_nepeng
  data_files:
  - split: test
    path:
    - "sa/sa_nepeng.csv"
- config_name: sa_spaeng
  data_files:
  - split: test
    path:
    - "sa/sa_spaeng.csv"
- config_name: sa_tameng
  data_files:
  - split: test
    path:
    - "sa/sa_tameng.csv"
- config_name: truthfulqa_araeng
  data_files:
  - split: test
    path:
    - "truthfulqa/truthfulqa_araeng.csv"
- config_name: truthfulqa_chieng
  data_files:
  - split: test
    path:
    - "truthfulqa/truthfulqa_chieng.csv"
- config_name: truthfulqa_hineng
  data_files:
  - split: test
    path:
    - "truthfulqa/truthfulqa_hineng.csv"
- config_name: truthfulqa_spaeng
  data_files:
  - split: test
    path:
    - "truthfulqa/truthfulqa_spaeng.csv"
license: apache-2.0
language:
- zh
- en
- es
- hi
- de
- nl
- fy
- fr
- ar
- bn
- mr
- ne
- ta
- ml
- gn
- ne

size_categories:
- 10K<n<100K

task_categories:
- text-generation
- question-answering
- translation
- text-classification

tags:
- code-mixing
- multilingual
- llm-evaluation
- benchmark
---
# ℹ️Dataset Card for CodeMixBench

## [EMNLP'25] [CodeMixBench: Evaluating Code-Mixing Capabilities of LLMs Across 18 Languages](https://arxiv.org/abs/2507.18791)

   <a href="https://github.com/Jeromeyluck/CodeMixBench" target="_blank">
      <img alt="Github" src="https://img.shields.io/badge/🐙-Github-blue" />
   </a>
        
  <a href="https://arxiv.org/abs/2507.18791" target="_blank">
      <img alt="Paper" src="https://img.shields.io/badge/📜-Paper-purple" />
   </a>
  <a href="https://2025.emnlp.org/" target="_blank">
      <img alt="EMNLP 2025" src="https://img.shields.io/badge/Proceedings-EMNLP2025-blue" />
   </a>



<!-- Provide a quick summary of the dataset. -->

Code-mixing is a linguistic phenomenon where multilingual speakers switch or mix two or more languages within a single utterance or conversation. 
To evaluate LLMs’ comprehension of multilingual code-mixed texts, we introduce CodeMixBench, a benchmark comprising eight tasks across 18 languages. 

![Statistics of 18 languages](./pics/18_languages.png)


## 🔎Dataset Details

Our benchmark comprises synthesized datasets targeting knowledge reasoning, 
mathematical reasoning, and truthfulness tasks, along with LID, POS, NER, SA, and MT tasks, 
which have been adapted from open-source studies. 


### CodeMixBench vs. Others

Previous benchmarks, such as GLUECoS and LinCE, primarily focus on traditional NLP tasks and are limited to a small number of languages. 
LinCE includes four language pairs and five NLP tasks: Language Identification(LID), 
Part of Speech (POS), Named Entity Recognition (NER), Sentiment Analysis (SA), and Machine Translation (MT). 
In contrast, GLUECoS covers two language pairs, lacks the MT task, but adds Question Answering (QA) and Natural Language Inference (NLI). 
Our review of recent codemixing studies indicates that research extends beyond the language pairs used in LinCE and GLUECoS. 
Therefore, we expanded to 16 language pairs and introduced tasks better suited for evaluating LLMs, 
such as Multi-Choice, Math, and Truthfulness, resulting in a total of eight tasks.

![language pairs and tasks](./pics/language_pairs.png)

### Statistics of Synthetic Datasets
For knowledge reasoning, we developed the code-mixed MMLU (CM-MMLU) based on the MMLU test set, 
featuring multiple-choice questions from 57 subjects to assess the model's comprehensive knowledge reasoning abilities. 
For mathematical reasoning, we created the code-mixed GSM8K (CM-GSM8K), derived from the GSM8K test set, 
which evaluates mathematical reasoning capabilities with each question including step-by-step solutions. 
For truthfulness assessment, we constructed the code-mixed TruthfulQA (CM-TruthfulQA) using 817 multiple-choice 
questions from the TruthfulQA test set. 

![Main evaluation results on CodeMixBench](./pics/statistics_synthesized_dataset.png)

### Statistics of Collected Datasets
We selected and reconstructed 30 datasets from existing open-source projects. To comprehensively evaluate the performance of large 
models on code-mixing, we aimed to encompass a diverse range of language families and tasks, prioritizing manually annotated datasets. 
Ultimately, we cover traditional NLP tasks such as Language Identification (LID), Named Entity Recognition (NER), 
Part-of-Speech tagging (POS), Sentiment Analysis(SA), and Machine Translation (MT), and cover 16 languages from seven language families: 
Germanic(en, de, nl, fy), Sino-Tibetan (zh, hok), Romance(es), Afro-Asiatic (msa, ea), Indo-Aryan (hi, bn, ne,mr), Dravidian (ta, ml), and Tupian (gn).

![Main evaluation results on CodeMixBench](./pics/statistics_collected_dataset.png)

### Experience Results
We evaluate three families of LLMs on CodeMixBench, revealing consistent underperformance across all models on code-mixing 
datasets involving language pairs from different language families. However, enhancements 
in training data size, model scale, post-training, and few-shot learning can improve LLM performance on code-mixing datasets.

![Main evaluation results on CodeMixBench](./pics/main_result_lineplot.png)

![Main evaluation results on CodeMixBench](./pics/other_result.png)



## 🚀Load CodeMixBench

   Taking the GSM8K task with mixed Chinese and English, gsm8k_chieng, as an example.
   
   ```python
   from datasets import load_dataset

   dataset_dict = load_dataset('CodeMixBench/CodeMixBench', data_files={'test': './gsm8k/gsm8k_chieng.csv'})
   ```

### 📍Dataset Sources

<!-- Provide the basic links for the dataset. -->

- **Repository:** https://github.com/Jeromeyluck/CodeMixBench/
- **Paper:** [CodeMixBench: Evaluating Code-Mixing Capabilities of LLMs Across 18 Languages](https://huggingface.co/papers/2507.18791)

## Setup

1. Follow these steps to set up your development environment:
   ```bash
   git clone git@github.com:Jeromeyluck/CodeMixBench.git
   cd CodeMixBench

   conda create -n CodeMixBench python=3.9
   conda activate CodeMixBench
   pip install -r requirements.txt
   ```
   
2. To launch an llm for testing:
   ```bash
   python ./test_model.py \
     --dataset lid_guaspa \
     --expid lid_guaspa_all_0shot \
     --model gpt-3.5-turbo \
     --shot 5 \
     --api sk-********************* \
     --url https://****************
   ```
   - `dataset`: select the dataset (e.g., `lid_gereng`, `lid_spaeng`, `ner_hineng`).
   - `expid`: define the ID of the test, the results file will be named after this ID.
   - `model`: the model you test. The default model is `gpt-3.5-turbo`.
   - `shot`: use for few-shot test (by default it will be `1`).
   - `api`: API Key (default key will be `OPENAI_API_KEY` defined in system path).
   - `url`: API function provider's URL.


## 🔗Citation

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->

**BibTeX:**

```
 @misc{yang2025codemixbenchevaluatingcodemixingcapabilities,
    title={CodeMixBench: Evaluating Code-Mixing Capabilities of LLMs Across 18 Languages}, 
    author={Yilun Yang and Yekun Chai},
    year={2025},
    eprint={2507.18791},
    archivePrefix={arXiv},
    primaryClass={cs.CL},
    url={https://arxiv.org/abs/2507.18791}, 
}
```