AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Perturb Your Data: Paraphrase-Guided Training Data Watermarking

Pranav Shetty, Mirazul Haque, Petr Babkin, Zhiqiang Ma, Xiaomo Liu, Manuela Veloso

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Text watermarking · 1 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
sample paraphrases with score-matched strategy instead of maximum Min-K%++ (ablation)SPECTRA · Wiki (member)p-value: 1.00 -> 4E-19Improved
sample paraphrases with score-matched strategy instead of maximum Min-K%++ (ablation)SPECTRA · HN (member)p-value: 3E-6 -> 3E-60Improved
sample paraphrases with score-matched strategy instead of random (ablation)SPECTRA · PeS2o (member)p-value: 1E-3 -> 2E-12Improved
sample paraphrases with score-matched strategy instead of random (ablation)SPECTRA · Wiki (non-member)p-value: 2E-5 -> 0.02Improved
use SPECTRA instead of STAMP watermarkingSPECTRA · PeS2o (member)p-value: 0.15 -> 2E-12Improved
use SPECTRA instead of LLM-DISPECTRA · Wiki (member)p-value: 0.02 -> 4E-19Improved
replace Min-K%++ with Loss in SPECTRA (ablation)SPECTRA · Wiki (non-member)p-value: 0.02 -> 6.2E-20Hurt
replace Min-K%++ with DC-PDD in SPECTRA (ablation)SPECTRA · Wiki (member)p-value: 4E-19 -> 1.0Hurt
use smaller scoring model Pythia-160m for ranking instead of OLMo-7b (ablation)SPECTRA · PeS2o paraphrase rankingSpearman rho: 0.826 -> 0.699Hurt
vary scoring model in the statistical test to OLMo-7b (ablation)SPECTRA · PeS2op-value: 2E-12 -> 3E-5No clear effect
increase number of samples to 100-150 (ablation)SPECTRA · all datasetsp-value: 100-150 samplesImproved
increase pretraining tokens beyond 3.5 billion (ablation)SPECTRA · all datasetsp-value: 3.5 billion tokensNo clear effect

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