OriginBlame: Record- and Token-Level Data Provenance for AI Training Datasets
Quick Answer
OriginBlame introduces a record- and token-level data provenance system that accurately tracks author contributions in AI training datasets, significantly reducing over-deletion from 101x to 1.3x.
Quick Take
Evaluations on 219,555 Wikipedia pages show a 42% improvement in unlearning efficiency for a 1.7B model, with minimal throughput overhead of 1.3-19.0%.
Key Points
- Record-level provenance reduces dataset-level over-deletion from 101x to 1.3x.
- Provenance-based forget sets improve unlearning by 42% compared to random methods.
- Integration adds 1.3-4.0% throughput overhead on HuggingFace datasets.
- Evaluated on 219,555 Wikipedia pages for robust performance metrics.
- Addresses practical challenges in data contributor removal requests.
Paper Resources
📖 Reader Mode
~2 min readAbstract:When a data contributor requests removal, model trainers face a practical gap: unlearning algorithms require a forget set, yet no tool can locate which training records belong to a given author. Existing provenance systems operate at file or dataset level, forcing catastrophic over-deletion. We present ob, a record- and token-level data provenance system that propagates author identity through data processing pipelines and resolves revocation requests into precise forget sets via deterministic queries. Evaluation on 219,555 Wikipedia pages demonstrates that record-level provenance eliminates dataset-level over-deletion (from 101x to 1.3x), while integration adds 1.3-4.0% throughput overhead (HuggingFace) and 2.1-19.0% (Datatrove) on wiki data. On a 1.7B model, provenance-based forget sets improve unlearning by 42% over random baselines.
| Comments: | 13 pages, 6 figures |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.13037 [cs.AI] |
| (or arXiv:2607.13037v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.13037 arXiv-issued DOI via DataCite |
Submission history
From: Haolin Xue [view email]
[v1]
Tue, 19 May 2026 10:18:33 UTC (81 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.