---
license: cc-by-nc-4.0
task_categories:
- multiple-choice
- translation
language:
- zh
tags:
- traditional chinese
- cantonese
- hong kong sar
- llm
- evaluation
pretty_name: HKMMLU
size_categories:
- 10K<n<100K
configs:
- config_name: basic_medical_science
  data_files:
  - split: test
    path: "data/multiple_choice/test/basic_medical_science.csv"
  - split: validation
    path: "data/multiple_choice/dev/basic_medical_science.csv"
- config_name: business_management
  data_files:
  - split: test
    path: "data/multiple_choice/test/business_management.csv"
  - split: validation
    path: "data/multiple_choice/dev/business_management.csv"
- config_name: chinese_conditions
  data_files:
  - split: test
    path: "data/multiple_choice/test/chinese_conditions.csv"
  - split: validation
    path: "data/multiple_choice/dev/chinese_conditions.csv"
- config_name: chinese_history
  data_files:
  - split: test
    path: "data/multiple_choice/test/chinese_history.csv"
  - split: validation
    path: "data/multiple_choice/dev/chinese_history.csv"
- config_name: chinese_world_culture
  data_files:
  - split: test
    path: "data/multiple_choice/test/chinese_world_culture.csv"
  - split: validation
    path: "data/multiple_choice/dev/chinese_world_culture.csv"
- config_name: chinese_world_geography
  data_files:
  - split: test
    path: "data/multiple_choice/test/chinese_world_geography.csv"
  - split: validation
    path: "data/multiple_choice/dev/chinese_world_geography.csv"
- config_name: clinical_psychology
  data_files:
  - split: test
    path: "data/multiple_choice/test/clinical_psychology.csv"
  - split: validation
    path: "data/multiple_choice/dev/clinical_psychology.csv"
- config_name: college_math
  data_files:
  - split: test
    path: "data/multiple_choice/test/college_math.csv"
  - split: validation
    path: "data/multiple_choice/dev/college_math.csv"
- config_name: culture_commonsense
  data_files:
  - split: test
    path: "data/multiple_choice/test/culture_commonsense.csv"
  - split: validation
    path: "data/multiple_choice/dev/culture_commonsense.csv"
- config_name: dentistry
  data_files:
  - split: test
    path: "data/multiple_choice/test/dentistry.csv"
  - split: validation
    path: "data/multiple_choice/dev/dentistry.csv"
- config_name: economics
  data_files:
  - split: test
    path: "data/multiple_choice/test/economics.csv"
  - split: validation
    path: "data/multiple_choice/dev/economics.csv"
- config_name: educational_psychology
  data_files:
  - split: test
    path: "data/multiple_choice/test/educational_psychology.csv"
  - split: validation
    path: "data/multiple_choice/dev/educational_psychology.csv"
- config_name: financial_analysis
  data_files:
  - split: test
    path: "data/multiple_choice/test/financial_analysis.csv"
  - split: validation
    path: "data/multiple_choice/dev/financial_analysis.csv"
- config_name: fire_science
  data_files:
  - split: test
    path: "data/multiple_choice/test/fire_science.csv"
  - split: validation
    path: "data/multiple_choice/dev/fire_science.csv"
- config_name: hk_celebrity
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_celebrity.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_celebrity.csv"
- config_name: hk_cultural_art
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_cultural_art.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_cultural_art.csv"
- config_name: hk_current_affairs
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_current_affairs.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_current_affairs.csv"
- config_name: hk_education
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_education.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_education.csv"
- config_name: hk_entertainment_industry_management
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_entertainment_industry_management.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_entertainment_industry_management.csv"
- config_name: hk_events_festivals
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_events_festivals.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_events_festivals.csv"
- config_name: hk_film_television
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_film_television.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_film_television.csv"
- config_name: hk_government_political_system
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_government_political_system.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_government_political_system.csv"
- config_name: hk_historical_development_geography
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_historical_development_geography.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_historical_development_geography.csv"
- config_name: hk_history
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_history.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_history.csv"
- config_name: hk_human_geography
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_human_geography.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_human_geography.csv"
- config_name: hk_law
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_law.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_law.csv"
- config_name: hk_life_knowledge
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_life_knowledge.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_life_knowledge.csv"
- config_name: hk_literature
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_literature.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_literature.csv"
- config_name: hk_livelihood
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_livelihood.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_livelihood.csv"
- config_name: hk_music
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_music.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_music.csv"
- config_name: hk_party_politics
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_party_politics.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_party_politics.csv"
- config_name: hk_physical_geography
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_physical_geography.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_physical_geography.csv"
- config_name: hk_politicians_events
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_politicians_events.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_politicians_events.csv"
- config_name: hk_society
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_society.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_society.csv"
- config_name: hk_transport_infrastructure
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_transport_infrastructure.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_transport_infrastructure.csv"
- config_name: hk_transportation
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_transportation.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_transportation.csv"
- config_name: hk_urban_regional_development
  data_files:
  - split: test
    path: "data/multiple_choice/test/hk_urban_regional_development.csv"
  - split: validation
    path: "data/multiple_choice/dev/hk_urban_regional_development.csv"
- config_name: hkdse_biology
  data_files:
  - split: test
    path: "data/multiple_choice/test/hkdse_biology.csv"
  - split: validation
    path: "data/multiple_choice/dev/hkdse_biology.csv"
- config_name: hkdse_business_accounting_finance
  data_files:
  - split: test
    path: "data/multiple_choice/test/hkdse_business_accounting_finance.csv"
  - split: validation
    path: "data/multiple_choice/dev/hkdse_business_accounting_finance.csv"
- config_name: hkdse_chemistry
  data_files:
  - split: test
    path: "data/multiple_choice/test/hkdse_chemistry.csv"
  - split: validation
    path: "data/multiple_choice/dev/hkdse_chemistry.csv"
- config_name: hkdse_economics
  data_files:
  - split: test
    path: "data/multiple_choice/test/hkdse_economics.csv"
  - split: validation
    path: "data/multiple_choice/dev/hkdse_economics.csv"
- config_name: hkdse_geography
  data_files:
  - split: test
    path: "data/multiple_choice/test/hkdse_geography.csv"
  - split: validation
    path: "data/multiple_choice/dev/hkdse_geography.csv"
- config_name: hkdse_information_and_communication_technology
  data_files:
  - split: test
    path: "data/multiple_choice/test/hkdse_information_and_communication_technology.csv"
  - split: validation
    path: "data/multiple_choice/dev/hkdse_information_and_communication_technology.csv"
- config_name: hkdse_mathematics
  data_files:
  - split: test
    path: "data/multiple_choice/test/hkdse_mathematics.csv"
  - split: validation
    path: "data/multiple_choice/dev/hkdse_mathematics.csv"
- config_name: hkdse_physics
  data_files:
  - split: test
    path: "data/multiple_choice/test/hkdse_physics.csv"
  - split: validation
    path: "data/multiple_choice/dev/hkdse_physics.csv"
- config_name: hkdse_tourism
  data_files:
  - split: test
    path: "data/multiple_choice/test/hkdse_tourism.csv"
  - split: validation
    path: "data/multiple_choice/dev/hkdse_tourism.csv"
- config_name: human_behavior
  data_files:
  - split: test
    path: "data/multiple_choice/test/human_behavior.csv"
  - split: validation
    path: "data/multiple_choice/dev/human_behavior.csv"
- config_name: insurance_studies
  data_files:
  - split: test
    path: "data/multiple_choice/test/insurance_studies.csv"
  - split: validation
    path: "data/multiple_choice/dev/insurance_studies.csv"
- config_name: logic_reasoning
  data_files:
  - split: test
    path: "data/multiple_choice/test/logic_reasoning.csv"
  - split: validation
    path: "data/multiple_choice/dev/logic_reasoning.csv"
- config_name: management_accounting
  data_files:
  - split: test
    path: "data/multiple_choice/test/management_accounting.csv"
  - split: validation
    path: "data/multiple_choice/dev/management_accounting.csv"
- config_name: marketing_management
  data_files:
  - split: test
    path: "data/multiple_choice/test/marketing_management.csv"
  - split: validation
    path: "data/multiple_choice/dev/marketing_management.csv"
- config_name: mechanical
  data_files:
  - split: test
    path: "data/multiple_choice/test/mechanical.csv"
  - split: validation
    path: "data/multiple_choice/dev/mechanical.csv"
- config_name: nautical_science
  data_files:
  - split: test
    path: "data/multiple_choice/test/nautical_science.csv"
  - split: validation
    path: "data/multiple_choice/dev/nautical_science.csv"
- config_name: optometry
  data_files:
  - split: test
    path: "data/multiple_choice/test/optometry.csv"
  - split: validation
    path: "data/multiple_choice/dev/optometry.csv"
- config_name: organic_chemistry
  data_files:
  - split: test
    path: "data/multiple_choice/test/organic_chemistry.csv"
  - split: validation
    path: "data/multiple_choice/dev/organic_chemistry.csv"
- config_name: pharmacology
  data_files:
  - split: test
    path: "data/multiple_choice/test/pharmacology.csv"
  - split: validation
    path: "data/multiple_choice/dev/pharmacology.csv"
- config_name: pharmacy
  data_files:
  - split: test
    path: "data/multiple_choice/test/pharmacy.csv"
  - split: validation
    path: "data/multiple_choice/dev/pharmacy.csv"
- config_name: physics
  data_files:
  - split: test
    path: "data/multiple_choice/test/physics.csv"
  - split: validation
    path: "data/multiple_choice/dev/physics.csv"
- config_name: science_technology_commonsense
  data_files:
  - split: test
    path: "data/multiple_choice/test/science_technology_commonsense.csv"
  - split: validation
    path: "data/multiple_choice/dev/science_technology_commonsense.csv"
- config_name: social_commonsense
  data_files:
  - split: test
    path: "data/multiple_choice/test/social_commonsense.csv"
  - split: validation
    path: "data/multiple_choice/dev/social_commonsense.csv"
- config_name: statistics_and_machine_learning
  data_files:
  - split: test
    path: "data/multiple_choice/test/statistics_and_machine_learning.csv"
  - split: validation
    path: "data/multiple_choice/dev/statistics_and_machine_learning.csv"
- config_name: technical
  data_files:
  - split: test
    path: "data/multiple_choice/test/technical.csv"
  - split: validation
    path: "data/multiple_choice/dev/technical.csv"
- config_name: traditional_chinese_medicine_clinical_medicine
  data_files:
  - split: test
    path: "data/multiple_choice/test/traditional_chinese_medicine_clinical_medicine.csv"
  - split: validation
    path: "data/multiple_choice/dev/traditional_chinese_medicine_clinical_medicine.csv"
- config_name: veterinary_pathology
  data_files:
  - split: test
    path: "data/multiple_choice/test/veterinary_pathology.csv"
  - split: validation
    path: "data/multiple_choice/dev/veterinary_pathology.csv"
- config_name: veterinary_pharmacology
  data_files:
  - split: test
    path: "data/multiple_choice/test/veterinary_pharmacology.csv"
  - split: validation
    path: "data/multiple_choice/dev/veterinary_pharmacology.csv"
- config_name: world_entertainment
  data_files:
  - split: test
    path: "data/multiple_choice/test/world_entertainment.csv"
  - split: validation
    path: "data/multiple_choice/dev/world_entertainment.csv"
---
## HKMMLU Dataset

| [**📖 Paper**](https://arxiv.org/abs/2505.02177) |[**🤗 Leaderboard**](https://huggingface.co/spaces/zhenghaozhu/llm-hkmmlu-leaderboard)｜

HKMMLU dataset is a multi-task language understanding dataset that evaluates both Hong Kong linguistic competence and socio-cultural knowledge. 

## Dataset Summary
The HKMMLU includes 26,698 multi-choice questions across 66 subjects, organized into four categories: Science, Technology, Engineering, and Mathematics (STEM), Social Sciences, Humanities, and Other. To evaluate the multilingual understanding ability of LLMs, it additionally comprises 88,292 Mandarin-Cantonese translation tasks. 

![Alt text](hkmmlu-architecture.jpg)

## 🏆 Mini-Leaderboard

Model performance on multi-choice tasks in HKMMLU using zero-shot prompting.

| Model                       |     HKMMLU       |
| --------------------------- | :--------------: |
| DeepSeek-V3-0324            |       76.6       |
| DeepSeek-V3                 |       74.7       |
| GPT-4o                      |       70.5       |
| Qwen2.5-72B-Instruct        |       69.0       |
| Llama-3-Taiwan-70B-Instruct |       67.5       |
| Claude 3.7 Sonnet           |       67.2       |
| Qwen2.5-14B-Instruct        |       62.2       |
| Yi-1.5-34B-Chat             |       60.1       |
| Mistral-Large-Instruct      |       60.0       |
| Llama-3-70B-Instruct        |       58.9       |
| Llama-3.1-70B-Instruct      |       58.7       |
| Gemma-2-27B-IT              |       57.1       |
| Qwen2.5-7B-Instruct         |       56.9       |
| Yi-1.5-9B-Chat              |       52.1       |
| Llama-3-Taiwan-8B-Instruct  |       51.3       |
| Qwen2.5-3B-Instruct         |       49.9       |
| GLM-4-9B-Chat               |       48.4       |
| Mistral-Small-Instruct      |       44.6       |
| Llama-3-8B-Instruct         |       44.1       |
| Gemma-2-2B-IT               |       41.4       |
| Llama-3.1-8B-Instruct       |       39.5       |


<!-- ## Updates

**2025-05-08:** Update translation tasks to 88,292 for quality enhancement. -->

## Load data

```python
from datasets import load_dataset
hkmmlu_hklaw = load_dataset(r"chuxuecao/HKMMLU", 'hk_law')
print(hkmmlu_hklaw)
print(hkmmlu_hklaw['test'][0])
```

Output

```
DatasetDict({
    test: Dataset({
        features: ['Question', 'A', 'B', 'C', 'D', 'Answer'],
        num_rows: 407
    })
    validation: Dataset({
        features: ['Question', 'A', 'B', 'C', 'D', 'Answer'],
        num_rows: 5
    })
})
{'Question': '以下哪一選項最準確描述制定《中華人民共和國香港特別行政區維護國家安全法》的法律依據？', 'A': '《中華人民共和國憲法》和《中華人民共和國香港特別行政區基本法》', 'B': '《中華人民共和國憲法》、《中華人民共和國香港特別行政區基本法》和《全國人民代表大會關於建立健全香港特別行政區維護國家安全的法律制度和執行機制的決定》', 'C': '《中華人民共和國憲法》、《中華人民共和國香港特別行政區基本法》和《中華人民共和國刑法》', 'D': '《全國人民代表大會關於建立健全香港特別行政區維護國家安全的法律制度和執行機制的決定》、《中華人民共和國刑法》和《中華人民共和國國家安全法》', 'Answer': 'B'}
```

## Load all data at once

```python
from datasets import load_dataset

multi_choice_task_list = ['basic_medical_science', 'business_management', 'chinese_conditions', 'chinese_history', 'chinese_world_culture', 'chinese_world_geography', 'clinical_psychology', 'college_math', 'culture_commonsense', 'dentistry', 'economics', 'educational_psychology', 'financial_analysis', 'fire_science', 'hk_celebrity', 'hk_cultural_art', 'hk_current_affairs', 'hk_education', 'hk_entertainment_industry_management', 'hk_events_festivals', 'hk_film_television', 'hk_government_political_system', 'hk_historical_development_geography', 'hk_history', 'hk_human_geography', 'hk_law', 'hk_life_knowledge', 'hk_literature', 'hk_livelihood', 'hk_music', 'hk_party_politics', 'hk_physical_geography', 'hk_politicians_events', 'hk_society', 'hk_transport_infrastructure', 'hk_transportation', 'hk_urban_regional_development', 'hkdse_biology', 'hkdse_business_accounting_finance', 'hkdse_chemistry', 'hkdse_economics', 'hkdse_geography', 'hkdse_information_and_communication_technology', 'hkdse_mathematics', 'hkdse_physics', 'hkdse_tourism', 'human_behavior', 'insurance_studies', 'logic_reasoning', 'management_accounting', 'marketing_management', 'mechanical', 'nautical_science', 'optometry', 'organic_chemistry', 'pharmacology', 'pharmacy', 'physics', 'science_technology_commonsense', 'social_commonsense', 'statistics_and_machine_learning', 'technical', 'traditional_chinese_medicine_clinical_medicine', 'veterinary_pathology', 'veterinary_pharmacology', 'world_entertainment']

hkmmlu_multichoice = {k: load_dataset(r"chuxuecao/HKMMLU", k) for k in multi_choice_task_list}

```

## Categories

```python
stem_list = ['basic_medical_science', 'dentistry', 'dse_biology', 'dse_chemistry', 
                'dse_information_and_communication_technology', 'dse_mathematics', 'dse_physics', 
                'college_math','engineering_math', 'linear_algebra', 'mechanical', 'optometry', 
                'organic_chemistry', 'pharmacology', 'pharmacy', 'physics', 'statistics_and_machine_learning', 
                'traditional_chinese_medicine_clinical_medicine', 'veterinary_pathology','veterinary_pharmacology']
socialsci_list = ['business_management', 'chinese_world_geography', 'clinical_psychology', 'dse_business_accounting_finance', 
                    'dse_economic', 'dse_geography', 'dse_travel', 'economics', 'educational_psychology', 'financial_analysis', 
                    'fire_science', 'hk_physical_geography', 'hk_transport_infrastructure', 'hk_urban_regional_development', 
                    'insurance_studies', 'management_accounting', 'marketing_management', 'nautical_science']
humanities_list = ['chinese_conditions', 'chinese_history', 'chinese_world_culture', 'hk_current_affairs', 'hk_education', 
                    'hk_entertainment_arts_culture', 'hk_entertainment_celebrity_culture', 'hk_government_political_system', 
                    'hk_historical_development_geography', 'hk_history', 'hk_human_geography', 'hk_law', 'hk_literature', 
                    'hk_party_politics', 'hk_politicians_events', 'hk_society', 'human_behavior', 'world_entertainment_culture']
other_list = ['culture_commonsense', 'hk_entertainment_industry_management', 'hk_events_festivals', 'hk_film_television', 
                'hk_life_knowledge', 'hk_music','hk_people_livelihood', 'hk_transportation', 'logic_reasoning', 
                'science_technology_commonsense', 'social_commonsense', 'technical']
```

## Citation
```
@misc{cao2025measuringhongkongmassive,
      title={Measuring Hong Kong Massive Multi-Task Language Understanding}, 
      author={Chuxue Cao and Zhenghao Zhu and Junqi Zhu and Guoying Lu and Siyu Peng and Juntao Dai and Weijie Shi and Sirui Han and Yike Guo},
      year={2025},
      eprint={2505.02177},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2505.02177}, 
}
```