AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Data Provenance Auditing of Fine-Tuned Large Language Models with a Text-Preserving Technique

Yanming Li, Cédric Eichler, Nicolas Anciaux, Alexandra Bensamoun, Lorena Gonzalez-Manzano, Seifeddine Ghozzi

ICML 2026 · Large language models › Instruction and fine-tuning · approach: Safety alignment after fine-tuning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
increase number of watermarked documents (ablation)invisible Unicode cue-reply watermark · Blog1k, Poems, Newsregurgitation rate: Mistral Blog1k p 0.060 -> 0.431 (30 -> 50 docs)Improved
increase amount of unmarked fine-tuning documents (ablation)invisible Unicode watermark · Blog1k, Poems, Newsregurgitation rate: Mistral News 0.74 at 3,000 docs vs about 0.25 at 1,000 and 9,000No clear effect
increase number of distinct watermarks in fine-tuning data (1 to 50) (ablation)invisible Unicode watermark · Blog1k, Poems, Newsregurgitation rate: Mistral about 0.5-0.7 at U=1 -> above 0.85 at U=50Improved
use cue-reply watermark with ranking test against counterfactualsMistral-7B, LLaMA-2-7B fine-tuning audit · Blog1k, Poems, NewsTPR at 0% FPR: 96.7% TPR at 0% FPR with k=1Improved
replace homoglyph watermark baseline with invisible cue-reply watermark under 50 unique watermarkshomoglyph watermarking (Wei et al., 2024) · Blog1k, Poems, News, 50 watermarksdetection significance: baseline Z-scores about -0.1 to -0.6 (p>0.2) vs regurgitation rates >0.8Improved
fine-tune GPT-OSS-20B instead of 7B models (ablation)multi-watermark setting · Appendix Dregurgitation rateImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy