# OriginBlame Proposes Record- and Token-Level Provenance to Locate an Author's Data Inside a Training Set

The method targets a gap that makes data-removal and unlearning requests impractical today: nobody can point to which training records belong to a given contributor.

- Published: 2026-07-16T05:44:01.303Z
- Canonical: https://polylog.news/ai/2026-07-16/originblame-proposes-record-and-token-level-provenance-to-lo
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv (OriginBlame)](https://arxiv.org/abs/2607.13037)

A paper introducing OriginBlame tackles an operational gap behind data-removal and unlearning requests. When a contributor asks that their data be removed, model trainers face an obstacle: machine-unlearning algorithms need a defined "forge…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-07-16/originblame-proposes-record-and-token-level-provenance-to-lo (subscription information: https://polylog.news/pricing).