HalluSearch offers a robust pipeline that combines retrieval-augmented verification with factual splitting to identify hallucinations in LLM outputs across 14 languages.

• Leverages search-enhanced RAG techniques to verify factual accuracy
• Performs fine-grained analysis to precisely locate fabricated text spans
• Demonstrates competitive performance in hallucination detection tasks
• Addresses critical security concerns related to AI-generated misinformation

This research is vital for securing AI systems by ensuring information integrity and preventing the spread of unreliable content in multilingual contexts.

HalluSearch at SemEval-2025 Task 3: A Search-Enhanced RAG Pipeline for Hallucination Detection