Open access
Aug 2026
Multi-SALLM: a multilingual security assessment of generated code
Multi-SALLM, a benchmarking framework designed to systematically evaluate Large Language Models’ ability to generate secure code, reveals three key findings: functional correctness and security are closely related but not equivalent, and sampling strategy is a critical risk factor.
Mohammed Latif Siddiq, Noshin Ulfat, Nishat Raihan et al.
· International Conference on... · 0 citations