--- configs: - config_name: amh data_files: - split: train path: amh/amh_all.csv - config_name: amh_fill_blank data_files: - split: train_1 path: amh/amh_fill_1.csv - split: train_2 path: amh/amh_fill_2.csv - split: train_3 path: amh/amh_fill_3.csv - split: valid path: amh/amh_fill_valid.csv - config_name: amh_choice_english data_files: - split: english_1 path: amh/amh_english_test_1.csv - split: english_2 path: amh/amh_english_test_2.csv - split: english_3 path: amh/amh_english_test_3.csv - split: english_4 path: amh/amh_english_test_4.csv - split: english_5 path: amh/amh_english_test_5.csv - config_name: translate_amh_choice_english data_files: - split: english_1 path: translate-test/amh/amh_english_test_1.csv - split: english_2 path: translate-test/amh/amh_english_test_2.csv - split: english_3 path: translate-test/amh/amh_english_test_3.csv - config_name: amh_choice_native data_files: - split: native_1 path: amh/amh_native_test_1.csv - split: native_2 path: amh/amh_native_test_2.csv - split: native_3 path: amh/amh_native_test_3.csv - split: native_4 path: amh/amh_native_test_4.csv - split: native_5 path: amh/amh_native_test_5.csv - config_name: translate_amh_choice_native data_files: - split: native_1 path: translate-test/amh/amh_native_test_1.csv - split: native_2 path: translate-test/amh/amh_native_test_2.csv - split: native_3 path: translate-test/amh/amh_native_test_3.csv - config_name: amh_generation data_files: - split: native path: amh/amh_meaining_generation_native.csv - split: english path: amh/amh_meaining_generation_english.csv - config_name: eng data_files: - split: train path: eng/eng_all.csv - config_name: eng_fill_blank data_files: - split: train_1 path: eng/eng_fill_1.csv - split: train_2 path: eng/eng_fill_2.csv - split: train_3 path: eng/eng_fill_3.csv - split: valid path: eng/eng_fill_valid.csv - config_name: eng_generation data_files: - split: native path: eng/eng_meaining_generation_native.csv - config_name: eng_choice_native data_files: - split: native_1 path: eng/eng_native_test_1.csv - split: native_2 path: eng/eng_native_test_2.csv - split: native_3 path: eng/eng_native_test_3.csv - config_name: eng_choice_native data_files: - split: native_1 path: eng/eng_native_test_1.csv - split: native_2 path: eng/eng_native_test_2.csv - split: native_3 path: eng/eng_native_test_3.csv - split: native_4 path: eng/eng_native_test_4.csv - split: native_5 path: eng/eng_native_test_5.csv - config_name: gez_fill_blank data_files: - split: train_1 path: geez/geez_fill_1.csv - split: train_2 path: geez/geez_fill_2.csv - split: train_3 path: geez/geez_fill_3.csv - split: valid path: geez/gez_fill_valid.csv - config_name: gez_choice_english data_files: - split: english_1 path: geez/geez_english_test_1.csv - split: english_2 path: geez/geez_english_test_2.csv - split: english_3 path: geez/geez_english_test_3.csv - split: english_4 path: geez/geez_english_test_4.csv - split: english_5 path: geez/geez_english_test_5.csv - config_name: gez_choice_native data_files: - split: native_1 path: geez/geez_native_test_1.csv - split: native_2 path: geez/geez_native_test_2.csv - split: native_3 path: geez/geez_native_test_3.csv - split: native_4 path: geez/geez_native_test_4.csv - split: native_5 path: geez/geez_native_test_5.csv - config_name: gez_generation data_files: - split: native path: geez/gez-native-description.csv - split: english path: geez/geez_meaining_generation_english.csv - config_name: orm data_files: - split: train path: orm/orm_all.csv - config_name: orm_choice_english data_files: - split: english_1 path: orm/orm_english_test_1.csv - split: english_2 path: orm/orm_english_test_2.csv - split: english_3 path: orm/orm_english_test_3.csv - split: english_4 path: orm/orm_english_test_4.csv - split: english_5 path: orm/orm_english_test_5.csv - config_name: translate_orm_choice_english data_files: - split: english_1 path: translate-test/orm/orm_english_test_1.csv - split: english_2 path: translate-test/orm/orm_english_test_2.csv - split: english_3 path: translate-test/orm/orm_english_test_3.csv - config_name: orm_choice_native data_files: - split: native_1 path: orm/orm_native_test_1.csv - split: native_2 path: orm/orm_native_test_2.csv - split: native_3 path: orm/orm_native_test_3.csv - split: native_4 path: orm/orm_native_test_4.csv - split: native_5 path: orm/orm_native_test_5.csv - config_name: translate_orm_choice_native data_files: - split: native_1 path: translate-test/orm/orm_native_test_1.csv - split: native_2 path: translate-test/orm/orm_native_test_2.csv - split: native_3 path: translate-test/orm/orm_native_test_3.csv - config_name: orm_generation data_files: - split: native path: orm/orm_meaining_generation_native.csv - split: english path: orm/orm_meaining_generation_english.csv - config_name: orm_fill_blank data_files: - split: train_1 path: orm/orm_fill_1.csv - split: train_2 path: orm/orm_fill_2.csv - split: train_3 path: orm/orm_fill_3.csv - split: valid path: orm/orm_fill_valid.csv - config_name: tir data_files: - split: train path: tir/tir_all.csv - config_name: tir_fill_blank data_files: - split: train_1 path: tir/tir_fill_1.csv - split: train_2 path: tir/tir_fill_2.csv - split: train_3 path: tir/tir_fill_3.csv - split: valid path: tir/tir_fill_valid.csv - config_name: tir_generation data_files: - split: native path: tir/tir_meaining_generation_native.csv - split: english path: tir/tir_meaining_generation_english.csv - config_name: tir_choice_english data_files: - split: english_1 path: tir/tir_english_test_1.csv - split: english_2 path: tir/tir_english_test_2.csv - split: english_3 path: tir/tir_english_test_3.csv - split: english_4 path: tir/tir_english_test_4.csv - split: english_5 path: tir/tir_english_test_5.csv - config_name: tir_choice_native data_files: - split: native_1 path: tir/tir_native_test_1.csv - split: native_2 path: tir/tir_native_test_2.csv - split: native_3 path: tir/tir_native_test_3.csv - split: native_4 path: tir/tir_native_test_4.csv - split: native_5 path: tir/tir_native_test_5.csv - config_name: translate_tir_choice_english data_files: - split: english_1 path: translate-test/tir/tir_english_test_1.csv - split: english_2 path: translate-test/tir/tir_english_test_2.csv - split: english_3 path: translate-test/tir/tir_english_test_3.csv - config_name: translate_tir_choice_native data_files: - split: native_1 path: translate-test/tir/tir_native_test_1.csv - split: native_2 path: translate-test/tir/tir_native_test_2.csv - split: native_3 path: translate-test/tir/tir_native_test_3.csv --- # ProverbEval: Benchmark for Evaluating LLMs on Low-Resource Proverbs This dataset accompanies the paper: **"ProverbEval: Exploring LLM Evaluation Challenges for Low-resource Language Understanding"** [ArXiv:2411.05049v3](https://arxiv.org/abs/2411.05049) ## Dataset Summary **ProverbEval** is a culturally grounded evaluation benchmark designed to assess the language understanding abilities of large language models (LLMs) in low-resource settings. It consists of tasks based on proverbs in five languages: - Amharic - Afaan Oromo - Tigrinya - Ge’ez - English The benchmark focuses on three tasks: - **Task 1**: Multiple Choice Meaning Matching - **Task 2**: Fill-in-the-Blank - **Task 3**: Proverb Generation ## Supported Tasks and Formats Each task is formatted to support multilingual and cross-lingual evaluation: ### Task 1: Meaning Multiple Choice - Input: A proverb in a given language. - Output: Select one correct meaning from four choices. - Format: Multiple-choice question with optional language variants for choices (native or English). ### Task 2: Fill-in-the-Blank - Input: A proverb with one word removed and four candidate words. - Output: Select the most suitable word to complete the proverb. - Format: Cloze-style multiple-choice. ### Task 3: Generation - Input: A detailed description of a proverb in English or the native language. - Output: The matching proverb generated in the target language. - Format: Text generation. ## Languages and Statistics ``` | Language | Task 1 | Task 2 | Task 3 | |--------------|--------|--------|--------| | Amharic | 483 | 494 | 484 | | Afaan Oromo | 502 | 493 | 502 | | Tigrinya | 380 | 503 | 380 | | Ge’ez | 434 | 429 | 434 | | English | 437 | 462 | 437 | ``` Note: The dataset focuses on **test sets only**. Few-shot examples are also included for Task 2. ## Data Structure Each example includes the following fields (depending on the task): - `language` - `task` - `prompt` - `choices` or `description` - `answer` or `target_proverb` - `prompt_type` (native or English) - `choice_language` (native or English) ## Usage You can load the dataset directly using the `datasets` library: ```python from datasets import load_dataset # Load the full dataset dataset = load_dataset("israel/ProverbEval") ``` ``` . ├── amh │ ├── amharic-fill_test.csv │ ├── amh_english_test_1.csv │ ├── amh_english_test_2.csv │ ├── amh_english_test_3.csv │ ├── amh_fill_1.csv │ ├── amh_fill_2.csv │ ├── amh_fill_3.csv │ ├── amh_meaining_generation_english.csv │ ├── amh_meaining_generation_native.csv │ ├── amh_native_test_1.csv │ ├── amh_native_test_2.csv │ └── amh_native_test_3.csv ├── eng │ ├── eng_fill_test.csv │ ├── eng_meaining_generation_native.csv │ ├── eng_native_test_1.csv │ ├── eng_native_test_2.csv │ └── eng_native_test_3.csv ├── geez │ ├── geez_english_test_1.csv │ ├── geez_english_test_2.csv │ ├── geez_english_test_3.csv │ ├── geez_fill_1.csv │ ├── geez_fill_2.csv │ ├── geez_fill_3.csv │ └── geez_meaining_generation_english.csv ├── orm │ ├── orm_english_test_1.csv │ ├── orm_english_test_2.csv │ ├── orm_english_test_3.csv │ ├── orm_fill_1.csv │ ├── orm_fill_2.csv │ ├── orm_fill_3.csv │ ├── orm_meaining_generation_english.csv │ ├── orm_meaining_generation_native.csv │ ├── orm_native_test_1.csv │ ├── orm_native_test_2.csv │ ├── orm_native_test_3.csv │ └── oromo_fill_test.csv └── tir ├── tir_fill_1.csv ├── tir_fill_2.csv └── tir_fill_3.csv ``` ``` @article{azime2024proverbeval, title={ProverbEval: Exploring LLM Evaluation Challenges for Low-resource Language Understanding}, author={Azime, Israel Abebe and Tonja, Atnafu Lambebo and Belay, Tadesse Destaw and Chanie, Yonas and Balcha, Bontu Fufa and Abadi, Negasi Haile and Ademtew, Henok Biadglign and Nerea, Mulubrhan Abebe and Yadeta, Debela Desalegn and Geremew, Derartu Dagne and others}, journal={arXiv preprint arXiv:2411.05049}, year={2024} } ```