최신 언어 생성 모델에서 발생하는 환각을 감지하기 위해 합성 데이터를 활용하는 방법 소개
Leveraging pre-trained models on hallucination detection and natural language inference to efficiently identify fluent overgeneration hallucinations in model-aware and model-agnostic settings.
Developing novel systems for the SemEval-2024 hallucination detection task by leveraging synthetic data generation and ensemble approaches to achieve strong performance.