CoNLL

Tonic AI released PrivacyBench, an open benchmark for de-identifying semi-structured data from work tools like email and Slack, which introduces novel metrics for evaluating synthesis quality. The ben…

Researchers at arXiv (arXiv:2605.23035) used sparse autoencoders to decompose GPT-2 XL and Llama-3.1-8B into 16K-32K interpretable features per layer, finding that semantic features alone recovered 94…