Setloop/What-Gradients-Add-to-Text-Leakage-in-Split-Language-Models-Counted-per-Token-and-per-Document

Paper A: reproduction and peer-review release This dataset contains the final paper, the evidence used for its three experiments, and the historical material explicitly discussed in the paper. Final paper The current branch includes the 2 October 2026 punctuation revision of the 26-page manuscript. Its scientific content and experimental evidence are unchanged. Tag v1.0 retains the frozen release and its original manuscript; the updated PDF and source package are… See the full description on the dataset page: https://huggingface.co/datasets/Setloop/What-Gradients-Add-to-Text-Leakage-in-Split-Language-Models-Counted-per-Token-and-per-Document.