Curated Turkish prompt-injection dataset by Enes Deniz, with paired benign and attack examples for LLM security and AI red/blue teaming.
A scalable two-stage news recommender that retrieves relevant candidates and reranks them using hybrid lexical and semantic features to optimize top-K recommendation quality.
Curated English dataset of 600 matched benign and attack pairs for agentic prompt-injection and trust-boundary testing.
GSMA telecom documentation dataset creation pipeline with hard negative generation for embedding training. Features concurrent LLM validation, semantic similarity ranking, and DVC-based reproducible data processing.
Add a description, image, and links to the hard-negatives topic page so that developers can more easily learn about it.
To associate your repository with the hard-negatives topic, visit your repo's landing page and select "manage topics."