AI Paper Atlas

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

Beyond Heuristic Tuning: Power-Calibrated LLM Watermarking

Xiaopu Wang, Zelin He, Chengyuan Liu, Runze Li

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
choose (gamma, delta) by power-calibrated optimization under KL budget instead of grid searchKGW watermark · C4, WinMax-C attack 10%, GPT-2largest -log(KL) at TPR>=0.95: 2.15 -> 2.70Improved
choose (gamma, delta) by power calibration instead of grid searchKGW watermark · C4, WinMax-C attack 10%, OPT-125Mlargest -log(KL) at TPR>=0.95: 1.86 -> 2.53Improved
choose (gamma, delta) by power calibration instead of DP (KL difference-in-probability)KGW watermark · C4, WinMax-C attack 10%, Pythia-160Mlargest -log(KL) at TPR>=0.95: 1.34 -> 2.70Improved
choose (gamma, delta) by power calibration instead of OPT-style log-perplexity optimizationKGW watermark · C4, WinMax-C attack 20%, GPT-2largest -log(KL) at TPR>=0.95: 2.01 -> 2.69Improved
choose the higher root gamma_h of the KL budget equation instead of the lower root (ablation)Power-calibrated watermark · Theory, alpha=0.05, delta=2, n=50KL at equal powerImproved
use WinMax-C detector instead of full-sequence z-score under local signal removalKGW watermark detection · signal-removal attackexpected signalImproved
theoretical gamma' formula versus observed green-token rateDirichlet-prior theory · C4/LFQA/Wikipedia with OPT, Pythia, GPT-2R2: R2 >= 0.98Improved
use KL divergence as proxy for text qualityKL distortion metric · LFQA quality metricsR2 vs BLEU: R2 0.566 (BLEU), 0.870 (ROUGE), 0.791 (BERTScore)Mixed

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