Grammar-Constrained Decoding Can Jailbreak LLMs into Generating Malicious Code

arXiv:2606.11817v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are increasingly used for code generation, raising concerns that they may be misused to produce malicious code. Meanwhile, Grammar-Constrained Decoding (GCD) has…

Read the full story at arXiv cs.AI ↗