Update README.md

README.md

@@ -314,141 +314,245 @@ dataset_info:
   download_size: 10010
   dataset_size: 12666
 ---
-
-# Dataset Card for Tokenization Robustness
-
 <!-- Provide a quick summary of the dataset. -->
 
-A comprehensive evaluation dataset for testing robustness of different tokenization strategies.
+<div align="center">
+  <img src="https://via.placeholder.com/800x200/4A90E2/FFFFFF?text=TokSuite" alt="TokSuite Logo" width="600"/>
+  
+  # Farsi Tokenizer Robustness Dataset
+  
+</div>
 
-## Dataset Details
+## Dataset Description
 
-### 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 Farsi (Persian) language multiple-choice text completion questions with various real-world perturbations that test tokenizer robustness.
 
-<!-- Provide a longer summary of what this dataset is. -->
+- **Curated by:** R3 Research Team
+- **Language(s):** Farsi/Persian (fa)
+- **License:** Creative Commons
 
-This dataset evaluates how robust language models are to different tokenization strategies and edge cases. It includes text completion questions with multiple choice answers designed to test various aspects of tokenization handling.
+### Dataset Summary
 
-- **Curated by:** R3
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** cc
+TokSuite addresses a fundamental challenge in language model research: understanding how tokenization choices impact model behavior in isolation. The Farsi subset specifically measures model performance on canonical questions and various perturbations including orthographic variations, diacritics, morphological challenges, and noise commonly encountered when processing Farsi text.
 
-### Dataset Sources [optional]
+**Key Features:**
+- 45 canonical questions covering general knowledge, geography, science, and language understanding
+- Multiple perturbation types reflecting real-world text variations in Farsi
+- Parallel structure with TokSuite benchmark (available in English, Turkish, Italian, Chinese)
+- Native speaker curation ensuring linguistic authenticity
 
-<!-- Provide the basic links for the dataset. -->
+### Supported Tasks
 
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
+- **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 Farsi text understanding
 
-## Uses
+### Languages
 
-<!-- Address questions around how the dataset is intended to be used. -->
+The dataset contains text in Farsi (Persian) written in Arabic script (language code: `pes_Arab` / `fa`).
 
-### Direct Use
+## Dataset Structure
 
-<!-- This section describes suitable use cases for the dataset. -->
+### Data Instances
+
+An example from the dataset:
+
+```json
+{
+  "question": "رنگ آسمان",
+  "choices": ["آبی است", "قرمز است", "سبز است", "زرد است"],
+  "answer": 0,
+  "answer_label": "A",
+  "split": "test",
+  "subcategories": "Cannonical",
+  "lang": "pes_Arab",
+  "second_lang": "The color of the sky is",
+  "coding_lang": "",
+  "notes": "The color of the sky is",
+  "id": "301",
+  "set_id": 301.0,
+  "variation_id": 1.0
+}
+```
+
+### Data Fields
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `question` | `string` | The question text in Farsi (Persian Arabic script) |
+| `choices` | `list[string]` | Four multiple-choice answer options in Farsi |
+| `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 (e.g., "Cannonical", "Diacritics", "Romanization", "Noise") |
+| `lang` | `string` | Language code (pes_Arab = Persian/Farsi in Arabic script) |
+| `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 300-344) |
+| `variation_id` | `float64` | Variation number within a question set |
+
+### Data Splits
+
+| Split | Number of Examples |
+|-------|-------------------|
+| test  | 45 question sets with multiple variations |
+
+All data is in the `test` split as this is an evaluation benchmark.
 
-[More Information Needed]
+## Dataset Creation
 
-### Out-of-Scope Use
+### Curation Rationale
 
-<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
+This dataset was created to:
+1. Systematically evaluate how different tokenization strategies handle Farsi text
+2. Measure robustness against real-world text perturbations specific to Persian
+3. Support research into tokenization's impact on language model behavior
+4. Provide standardized benchmarks for Farsi language models
 
-[More Information Needed]
+The questions were designed to be straightforward with high baseline accuracy, allowing researchers to cleanly measure performance degradation when perturbations are applied.
 
-## Dataset Structure
+### Source Data
 
-<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
+#### Data Collection and Processing
 
-The dataset contains multiple-choice questions with associated metadata about tokenization types and categories.
+- **Canonical Questions**: 40 baseline questions in English were created covering general knowledge topics
+- **Translation**: Native Farsi speakers translated questions to Persian
+- **Perturbations**: Each question underwent targeted perturbations designed to reflect morphological and orthographic characteristics of Farsi
+- **Validation**: Model-in-the-loop process ensured high baseline accuracy across 14 different tokenizers
 
-## Dataset Creation
+#### Perturbation Categories
 
-### Curation Rationale
+1. **Orthographic Perturbations**
+   - Script variations (traditional vs. simplified)
+   - Romanization (Finglish - Farsi in Latin script)
+   - Homoglyphs (visually similar characters with different Unicode)
+   - Zero-width characters and spacing irregularities
 
-<!-- Motivation for the creation of this dataset. -->
+2. **Diacritics**
+   - Optional short vowels (fatha /a/, kasra /e/, damma /o/)
+   - Common accent errors
 
-[More Information Needed]
+3. **Morphological Challenges**
+   - Contractions and compound words
+   - Inflectional variations
+   - Case marking and derivations
 
-### Source Data
-
-<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
+4. **Input Medium Challenges**
+   - Non-native keyboard typing (e.g., typing Farsi with English keyboard)
+   - Systematic character substitutions
 
-#### Data Collection and Processing
+5. **Noise**
+   - Typos and character-level errors
+   - OCR-like errors
+   - Character deletion/permutation
+   - Spacing inconsistencies
 
-<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
+6. **Linguistic Variety**
+   - Code-switching
+   - Dialectal variations
+   - Historical spelling variations
 
-[More Information Needed]
+7. **Structural Elements**
+   - Unicode-based formatting
+   - Stylistic variations
 
 #### Who are the source data producers?
 
-<!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->
+Native Farsi speakers curated and validated all questions and perturbations. The TokSuite research team at R3 designed the overall benchmark framework.
 
-[More Information Needed]
+### Annotations
 
-### Annotations [optional]
+#### Annotation process
 
-<!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
+Questions were manually created and translated by native speakers. Each perturbation was carefully designed to reflect authentic variations encountered in real-world Farsi text processing.
 
-#### Annotation process
+#### Who are the annotators?
 
-<!-- This section describes the annotation process such as annotation tools used in the process, the amount of data annotated, annotation guidelines provided to the annotators, interannotator statistics, annotation validation, etc. -->
+Native Farsi speakers with expertise in linguistics and NLP, working as part of the TokSuite project.
 
-[More Information Needed]
+### Personal and Sensitive Information
 
-#### Who are the annotators?
+The dataset contains only general knowledge questions and does not include any personal or sensitive information.
 
-<!-- This section describes the people or systems who created the annotations. -->
+## Considerations for Using the Data
 
-[More Information Needed]
+### Social Impact of Dataset
 
-#### Personal and Sensitive Information
+This dataset contributes to improving language technology for Farsi speakers by:
+- Enabling better understanding of tokenization challenges in Persian
+- Supporting development of more robust multilingual models
+- Providing standardized evaluation for Farsi NLP research
 
-<!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. -->
+### Discussion of Biases
 
-[More Information Needed]
+- **Language variety**: The dataset uses Modern Standard Persian and may not fully represent dialectal variations
+- **Script focus**: Only Arabic script is used; romanized versions are included as perturbations
+- **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
 
-## Bias, Risks, and Limitations
+### Other Known Limitations
 
-<!-- This section is meant to convey both technical and sociotechnical 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 Farsi's characteristics and findings may not generalize to all languages
+- Models evaluated were trained at ~1B parameters; results may differ at larger scales
 
-The dataset focuses primarily on English text and may not generalize to other languages or tokenization schemes not covered in the evaluation.
+## Additional Information
 
-### Recommendations
+### Dataset Curators
 
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+The dataset was curated by the TokSuite research team at R3.
 
-Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations.
+### Licensing Information
 
-## Citation [optional]
+Creative Commons license (cc)
 
-<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
+### Citation Information
 
-**BibTeX:**
+If you use this dataset in your research, please cite the TokSuite paper:
 
-[More Information Needed]
+```bibtex
+@inproceedings{toksuite2026,
+  title={TokSuite: Measuring the Impact of Tokenizer Choice on Language Model Behavior},
+  author={Anonymous},
+  booktitle={Under review as a conference paper at ICLR 2026},
+  year={2026},
+  url={https://openreview.net/pdf?id=iExjy56t3o}
+}
+```
 
-**APA:**
+**Paper**: [TokSuite: Measuring the Impact of Tokenizer Choice on Language Model Behavior](https://openreview.net/pdf?id=iExjy56t3o)
 
-[More Information Needed]
+### Contributions
 
-## Glossary [optional]
+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
 
-<!-- If relevant, include terms and calculations in this section that can help readers understand the dataset or dataset card. -->
+### Key Findings from TokSuite (Farsi-Specific)
 
-[More Information Needed]
+Based on the TokSuite paper findings for Farsi:
 
-## More Information [optional]
+- **Byte-level tokenizers** (like ByT5) demonstrate greater robustness to Farsi perturbations despite higher computational costs
+- **Multilingual tokenizers** with insufficient Farsi representation show significant performance degradation
+- **Diacritics and Unicode formatting** present challenges across nearly all tokenization strategies
+- **Average subword fertility** for Farsi ranges from 1.36 to 7.74 depending on tokenizer choice
+- **Performance drops** on perturbed Farsi text are notably higher (avg. 0.45) compared to English equivalents (avg. 0.11)
+- **Noise perturbations** cause more severe degradation in Farsi (0.22) than in English (0.15)
 
-[More Information Needed]
+### Contact
 
-## Dataset Card Authors [optional]
+For questions or issues related to this dataset, please refer to the TokSuite project or contact the authors through the paper submission system.
+
+---
 
-[More Information Needed]
+<div align="center">
+  
+**Part of the [TokSuite Project](https://openreview.net/pdf?id=iExjy56t3o)**
 
-## Dataset Card Contact
+*Understanding Tokenization's Role in Language Model Behavior*
 
-[More Information Needed]
+</div>