LLM agents inject CWEs into student-authored code to generate personalized security examples; in a 71-student deployment, participants rated them more relevant than textbook cases but quantitative differences remained limited.
VulScribeR: Exploring RAG-based vulnerability augmentation with LLMs,
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CR 2years
2026 2representative citing papers
Abliteration eliminates refusal in Qwen2.5-Coder models for CWE-89 prompts with syntactic validity above 93 percent, while post-edit injection success scales with model size from 25-48 percent at 3B to 88-97 percent at 14B.
citing papers explorer
-
Towards Personalizing Secure Programming Education with LLM-Injected Vulnerabilities
LLM agents inject CWEs into student-authored code to generate personalized security examples; in a 71-student deployment, participants rated them more relevant than textbook cases but quantitative differences remained limited.
-
Willing but Unable: Separating Refusal from Capability in Code LLMs via Abliteration
Abliteration eliminates refusal in Qwen2.5-Coder models for CWE-89 prompts with syntactic validity above 93 percent, while post-edit injection success scales with model size from 25-48 percent at 3B to 88-97 percent at 14B.