AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio

Georgios Milis, Yubin Qin, Yihan Wu, Heng Huang

ICML 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Prediction-powered and conformal inference · 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 vocabulary clustering (cluster-level watermark) to KGWKGW on Moshi (h=0) · Moshi conversational, identitymean -log p-value: 8.51 -> 42.47Improved
add vocabulary clustering to KGW under noise attackKGW on Moshi (h=0) · Moshi conversational, noise attackmean -log p-value: 2.23 -> 20.59Improved
add vocabulary clustering to KGW under croppingKGW on Moshi (h=0) · Moshi conversational, crop attackmean -log p-value: 1.51 -> 16.48Improved
add vocabulary clustering versus codec finetuning (WMAR)Moshi watermark · Moshi conversational, identitymean -log p-value: 17.44 -> 42.47Improved
add vocabulary clustering to KGW on LibriSpeech promptsKGW on Moshi (h=0) · Moshi LibriSpeech, identitymean -log p-value: 3.61 -> 22.77Improved
add vocabulary clustering to KGW, audio qualityKGW on Moshi (h=0) · Moshi conversational, h=0NISQA MOS: 3.46 -> 3.53No clear effect
compare clustering to WMAR finetuning on audio qualityMoshi watermark · Moshi conversational, h=0FAD VGGish: 0.407 -> 0.133Improved
add clustering to KGW on text-to-speechCosyVoice3 · CosyVoice3mean -log p-value: 1.564 -> 13.927Improved
add clustering to KGW on text-to-speechSpark-TTS · Spark-TTSmean -log p-value: 9.237 -> 17.806Improved
add clustering to KGW, speech recognition errorsCosyVoice3 · CosyVoice3WER: 0.0586 -> 0.0519Improved
add clustering to KGW, FAD VGGishSpark-TTS · Spark-TTSFAD VGGish: 0.2221 -> 0.3506Hurt
use Leiden instead of Louvain community detection (ablation)vocabulary clustering · Moshiwatermark detectabilityImproved

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