Turkish tool-calling dataset
Training and evaluation data for models that call tools from Turkish requests.
What it is
Tool calling is a language model's ability to pick the right tool for a request and call it with valid arguments. Most training and evaluation data for this is in English.
We are building a pipeline that generates examples made of a Turkish request, a set of function definitions and the expected call. Each example is checked for structural and argument correctness before it enters the dataset.
The dataset will be published under the Ilynt Labs organization on Hugging Face.
Why it exists
Building agents and automation for Turkish-speaking users requires models that call tools reliably from Turkish input. Measuring and improving that needs Turkish-specific data.
Who it is for
Developers and researchers who train, fine-tune or evaluate Turkish language models and agent systems.