Datasets:

toksuite
/

tokenizer_robustness_completion_turkish

Tasks:

Multiple Choice

Modalities:

Formats:

Languages:

Size:

ArXiv:

Tags:

Libraries:

License:

Dataset card Data Studio Files Files and versions

Malikeh1375 commited on 26 days ago

Commit

ed39cc2

·

verified ·

1 Parent(s): ff2865f

Update README.md

Files changed (1) hide show

README.md +1 -0

README.md CHANGED Viewed

@@ -2212,6 +2212,7 @@ This dataset is part of **TokSuite**, a comprehensive benchmark designed to meas
 ### Dataset Summary
 TokSuite addresses a fundamental challenge in language model research: understanding how tokenization choices impact model behavior in isolation. The Turkish subset specifically measures model performance on canonical questions and various perturbations.
 **Key Features:**
 - 40 canonical questions covering general knowledge, geography, science, and language understanding
 - Multiple perturbation types reflecting real-world text variations in Turkish

 ### Dataset Summary
 TokSuite addresses a fundamental challenge in language model research: understanding how tokenization choices impact model behavior in isolation. The Turkish subset specifically measures model performance on canonical questions and various perturbations.
 **Key Features:**
 - 40 canonical questions covering general knowledge, geography, science, and language understanding
 - Multiple perturbation types reflecting real-world text variations in Turkish