Language Model Attribution

Language model attribution focuses on explaining the origins and justifications behind the outputs of large language models (LLMs), aiming to improve their reliability and trustworthiness. Current research emphasizes unifying different attribution methods, such as citing training data and generating corroborating evidence, and developing benchmarks to evaluate these methods across various tasks, including question answering and text generation. This work is crucial for building more responsible and accountable LLMs, addressing concerns about hallucinations and biases, and enabling wider adoption in high-stakes applications like legal and medical domains.