r-three/tokenizer_robustness_completion_math

Dataset Card for Tokenization Robustness Math

TokSuite Benchmark ({LANGUAGE_NAME} Collection)

Dataset Description

This dataset is part of TokSuite, a comprehensive benchmark designed to measure how different tokenization strategies affect language model performance and robustness. This specific subset contains {LANGUAGE_NAME} language multiple-choice text completion questions with various real-world perturbations that test tokenizer robustness.

• Curated by: R3 Research Team
• Language(s): {LANGUAGE_NAME} ({LANGUAGE_CODE})
• License: MIT License

Dataset Summary

TokSuite addresses a fundamental challenge in language model research: understanding how tokenization choices impact model behavior in isolation. The {LANGUAGE_NAME} subset specifically measures model performance on canonical questions and various perturbations including {LIST_KEY_PERTURBATION_TYPES}.

Key Features:

• {NUM_CANONICAL_QUESTIONS} canonical questions covering {TOPIC_AREAS}
• Multiple perturbation types reflecting real-world text variations in {LANGUAGE_NAME}
• Parallel structure with TokSuite benchmark (available in English, Turkish, Italian, Chinese, Farsi)
• Native speaker curation ensuring linguistic authenticity

Supported Tasks

• Multiple-Choice Question Answering: Text completion format with 4 answer choices
• Tokenizer Robustness Evaluation: Measuring performance degradation under various text perturbations
• Multilingual NLP Benchmarking: Evaluating language models on {LANGUAGE_NAME} text understanding

Languages

The dataset contains text in {LANGUAGE_NAME} written in {SCRIPT_NAME} (language code: {LANGUAGE_CODE_FULL}).

Dataset Structure

Data Instances

An example from the dataset:

{
  "question": "{EXAMPLE_QUESTION}",
  "choices": ["{CHOICE_A}", "{CHOICE_B}", "{CHOICE_C}", "{CHOICE_D}"],
  "answer": {ANSWER_INDEX},
  "answer_label": "{ANSWER_LABEL}",
  "split": "test",
  "subcategories": "{SUBCATEGORY}",
  "lang": "{LANGUAGE_CODE_FULL}",
  "second_lang": "{ENGLISH_TRANSLATION}",
  "coding_lang": "",
  "notes": "{NOTES}",
  "id": "{ID}",
  "set_id": {SET_ID},
  "variation_id": {VARIATION_ID}
}

Data Fields

Field	Type	Description
question	string	The question text in {LANGUAGE_NAME} ({SCRIPT_DESCRIPTION})
choices	list[string]	Four multiple-choice answer options in {LANGUAGE_NAME}
answer	int64	Index of the correct answer (0-3)
answer_label	string	Letter label of the correct answer (A, B, C, or D)
split	string	Dataset split identifier (all entries are "test")
subcategories	string	Perturbation category
lang	string	Language code ({LANGUAGE_CODE_FULL} = {LANGUAGE_DESCRIPTION})
second_lang	string	English translation or description of the question
coding_lang	string	Not applicable for this dataset (empty string)
notes	string	Additional context about the question or perturbation type
id	string	Unique question identifier
set_id	float64	Question set grouping identifier (ranges from {ID_RANGE_START}-{ID_RANGE_END})
variation_id	float64	Variation number within a question set

Dataset Creation

Curation Rationale

This dataset was created to:

1. Systematically evaluate how different tokenization strategies handle {LANGUAGE_NAME} text
2. Measure robustness against real-world text perturbations specific to {LANGUAGE_NAME} language
3. Support research into tokenization's impact on language model behavior
4. Provide standardized benchmarks for {LANGUAGE_NAME} language models

The questions were designed to be straightforward with high baseline accuracy, allowing researchers to cleanly measure performance degradation when perturbations are applied.

Source Data

Data Collection and Processing

• Canonical Questions: {NUM_BASE_QUESTIONS} baseline questions in English were created covering general knowledge topics
• Translation: Native {LANGUAGE_NAME} speakers translated questions to {LANGUAGE_NAME}
• Perturbations: Each question underwent targeted perturbations designed to reflect {LINGUISTIC_CHARACTERISTICS}
• Validation: Model-in-the-loop process ensured high baseline accuracy across 14 different tokenizers

Perturbation Categories

1. Canonical {DESCRIPTION_OF_CANONICAL}

2. {PERTURBATION_NAME_1} {DESCRIPTION_1}

3. {PERTURBATION_NAME_2} {DESCRIPTION_2}

4. {PERTURBATION_NAME_3} {DESCRIPTION_3}

5. {PERTURBATION_NAME_4} {DESCRIPTION_4}

6. {PERTURBATION_NAME_5} {DESCRIPTION_5}

7. {PERTURBATION_NAME_6} {DESCRIPTION_6}

8. {PERTURBATION_NAME_7} {DESCRIPTION_7}

Model Performance Comparison

model_name	canonical	{PERTURBATION_COL_1}	{PERTURBATION_COL_2}	{PERTURBATION_COL_3}	{PERTURBATION_COL_4}	{PERTURBATION_COL_5}	{PERTURBATION_COL_6}	{PERTURBATION_COL_7}
Aya	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
BLOOM	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
ByT5	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
Comma	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
GPT-2	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
GPT-4o	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
Gemma-2	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
Llama-3.2	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
Phi-3	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
Qwen-3	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
Tekken	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
TokenMonster	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
XGLM	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}
mBERT	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}	{VAL}

Who are the source data producers?

Native {LANGUAGE_NAME} speakers curated and validated all questions and perturbations. The TokSuite research team at R3 designed the overall benchmark framework.

Annotations

Annotation process

Questions were manually created and translated by native speakers. Each perturbation was carefully designed to reflect authentic variations encountered in real-world {LANGUAGE_NAME} text processing.

Who are the annotators?

Native {LANGUAGE_NAME} speakers with expertise in linguistics and NLP, working as part of the TokSuite project.

Personal and Sensitive Information

The dataset contains only general knowledge questions and does not include any personal or sensitive information.

Considerations for Using the Data

Social Impact of Dataset

This dataset contributes to improving language technology for {LANGUAGE_NAME} speakers by:

• Enabling better understanding of tokenization challenges in {LANGUAGE_NAME}
• Supporting development of more robust multilingual models
• Providing standardized evaluation for {LANGUAGE_NAME} NLP research

Discussion of Biases

• Language variety: The dataset uses {STANDARD_VARIETY} and may not fully represent dialectal variations
• Script focus: {SCRIPT_LIMITATIONS_DESCRIPTION}
• Domain coverage: Questions focus on general knowledge and may not represent domain-specific language use
• Question simplicity: Designed for high baseline accuracy, which may not reflect real-world task complexity

Other Known Limitations

• Relatively small dataset size (designed for evaluation, not training)
• Focus on multiple-choice format may not capture all aspects of language understanding
• Perturbations are specific to {LANGUAGE_NAME}'s characteristics and findings may not generalize to all languages
• Models evaluated were trained at ~1B parameters; results may differ at larger scales

Additional Information

Dataset Curators

The dataset was curated by the TokSuite research team at R3.

Licensing Information

MIT license

Citation Information

If you use this dataset in your research, please cite the TokSuite paper:

@inproceedings{toksuite2026,
  title={TokSuite: Measuring the Impact of Tokenizer Choice on Language Model Behavior},
  author={Altıntaş, Gül Sena and Ehghaghi, Malikeh and Lester, Brian and Liu, Fengyuan and Zhao, Wanru and Ciccone, Marco and Raffel, Colin},
  booktitle={Preprint.},
  year={2026},
  url={TBD}
}

Paper: TokSuite: Measuring the Impact of Tokenizer Choice on Language Model Behavior

Contributions

This dataset is part of TokSuite, which includes:

• 14 language models with identical architectures but different tokenizers
• Multilingual benchmark datasets (English, Turkish, Italian, Farsi, Chinese)
• Comprehensive analysis of tokenization's impact on model behavior

Contact

For questions or issues related to this dataset, please refer to the TokSuite project or contact the authors through the paper submission system.

---

Part of the TokSuite Project

Understanding Tokenization's Role in Language Model Behavior