Paper: https://arxiv.org/abs/2506.23394 Official evaluation dataset for Tucan models - a series of Bulgarian language models fine-tuned for function calling and tool use.

📊 Dataset Overview

This dataset contains 120 curated samples designed to evaluate function-calling capabilities in Bulgarian language models. Each sample includes:

User queries in Bulgarian Available functions with their parameters Expected behavior (function calls or… See the full description on the dataset page: https://huggingface.co/datasets/llm-bg/Tucan-BG-Eval-v1.0.