AI Paper Atlas

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

Stability-Aware Feature Design for Robust Watermark Detection in Machine-Generated Text

Sina Mansouri, Mohit Marvania, Abolfazl Safikhani

ICML 2026 · Large language models › Language, safety and interpretability · approach: Text watermarking · 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
add z-score moments (6-D) to global z-score (ablation)global z-score detector · PG-19 1500 tokens D1AUC (%): 74.2 -> 83.2Improved
add lag-1/lag-2 autocorrelations (8-D) (ablation)z-score moments detector · PG-19 1500 tokens D1AUC (%): 83.2 -> 85.0Improved
add run-length statistics (14-D) (ablation)autocorrelation detector · PG-19 1500 tokens D1AUC (%): 85.0 -> 87.3Improved
add run-frequency features (full 20-D) (ablation)run-length detector · PG-19 1500 tokens D8AUC (%): 77.2 -> 78.6Improved
add paraphrase stability score (PSS) to static features (ablation)static features detector · PG-19 1500 tokens D1AUC (%): 88.1 -> 96.1Improved
add PSS to static features (ablation)static features detector · PG-19 1500 tokens D8AUC (%): 78.6 -> 91.2Improved
replace global z-score with WinMax local maximumglobal z-score · PG-19 1500 tokens D1AUC (%): 74.2 -> 82.1Improved
use PSS + Static instead of deep-learning detector RADARRADAR · PG-19 1500 tokens D8AUC (%): 40.8 -> 91.2Improved
increase watermark greenlist ratio gamma from 0.25 to 0.50PSS + Static · PG-19 1500 tokens D8AUC (%): 91.2 -> 94.2Improved
train on PG-19 and test on CNN/DM instead of in-domainPSS + Static · PG-19 to CNN/DM D1AUC (%): 96.1 -> 88.6Hurt
change paraphraser to Gemma for universal classifierPSS + Static universal classifier · PG-19 D1AUC (%): 96.1 -> 89.1Hurt
change window/stride across 40-60 and 5-15 (ablation)PSS + Static · PG-19 1500 tokens mean AUC D1-D8AUC (%): 99.0 vs 95.95No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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