From Language to Action in Arabic: Reliable Structured Tool Calling via Data-Centric Fine-Tuning
Abstract
Function-calling language models are essential for agentic AI systems that translate natural language into executable structured actions, yet existing models exhibit severe structural instability when applied to Arabic. We present AISA-AR-FunctionCall, a production-oriented Arabic function-calling framework built on a 270M-parameter FunctionGemma backbone and trained through systematic dataset auditing, schema repair, tool-aware prompt restructuring, and full-parameter supervised fine-tuning. On a held-out test set, fine-tuning reduces parse failures from 87\% to below 1\%, improves function name accuracy by more than eightfold, and substantially enhances argument alignment across dialects and domains. Error analysis reveals a transition from structural collapse to semantic misalignment, suggesting that serialization stability and decision-level reasoning are separable challenges. We further explore a reasoning-augmented LoRA variant that introduces explicit intermediate reasoning prior to tool invocation. All datasets and models are publicly released under the AISA framework.
Cite
@article{arxiv.2603.16901,
title = {From Language to Action in Arabic: Reliable Structured Tool Calling via Data-Centric Fine-Tuning},
author = {Omer Nacar and Deema Alquffari and Saleh Alsharideh and Adeem AlOtaibi and Abdulaziz Alabdulkarim and Leen Alhazmi and Nada Alomar and Wareef Alzubaidi and Nada Alsultan and Ahmed Alrabghi and Demah Alhoshan and Rana Alsayyari and Hamed Alruwaili and Albaraa Jaafar and Khaled Alusmani and Abdulaziz Alsohimy and Munirah Alsubaie and Shahd Aldukhayil and Arwa Alali and Yazeed BinShihah and Razan Alsulaymi and Nourah Alhumaid and Razan Abdulsalam and Reem Alamoudi and Mohammed Alkhalifa},
journal= {arXiv preprint arXiv:2603.16901},
year = {2026}
}