Work

Mitigating Security Vulnerabilities in Open-Source Large Language Models via a Supervisor Agent Architecture

Year: 2026

Type: conference-paper

Abstract: This study proposes an independent Supervisor Agent approach based on a Turkish safety classifier to protect open-source large language models against harmful prompts. The proposed TurkGuard model classifies user prompts as Safe or Unsafe before they are forwarded to the main model and provides an additional security layer by blocking harmful prompts. The study addresses large language model security threats such as Prompt Injection and Jailbreaking and discusses the limitations of existing English-oriented safety solutions due to the morphological characteristics of Turkish. Experimental results show that the fine-tuned small-scale Qwen3.5-based TurkGuard model can provide high attack detection performance and a low false alarm rate for Turkish prompts.

Source: International Conference Computational Linguistics in Bulgaria (CLIB)

DOI: 10.47810/clib.26.03

Authors: Öner Aytaş, Mehmet Bayram, Tuğçe Şen, Banu Diri, Göksel Biricik

Institutions: Işık University, Yıldız Technical University

Language: English

Topic: Adversarial Robustness in Machine Learning

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: bronze

OpenAlex ID: W7223571006

FWCI:

Cites:

Cited by:

Related to: 10