#codevulnerability
White‑Basilisk, a 200 million‑parameter AI model, can scan whole code repositories in a single pass, handling far longer sequences than typical LLMs while staying lightweight. https://getnews.me/white-basilisk-sets-new-benchmark-for-code-vulnerability-detection/ #whitebasilisk #codevulnerability
September 27, 2025 at 3:06 AM
LLaVul, a multimodal LLM that answers security questions about code, outperforms CodeBERT and GPT‑4‑code on a QA dataset, per ICSC 2025. Read more: https://getnews.me/llavul-multimodal-llm-boosts-interpretable-code-vulnerability-reasoning/ #llavul #codevulnerability
September 24, 2025 at 9:11 PM