Search papers, labs, and topics across Lattice.
3
0
5
3
Cross-lingual fairness gaps in language model watermarking are not just language-specific but are fundamentally tied to the structural properties of language families.
Watermarking, intended to govern AI content, is held to a lower fairness standard than the AI models it's meant to regulate, potentially leading to biased content authentication.
Genomic language models memorize training data, raising privacy concerns, and this study shows that no single memorization attack can fully capture the risk, necessitating a multi-vector approach to auditing.
