AI Paper Atlas

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

Analyzing and Evaluating Unbiased Language Model Watermark

Yihan Wu, Xuehao Cui, Ruibo Chen, Heng Huang

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Text watermarking · 6 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
raise KGW bias strength delta from 0.5 to 2.0KGW · WMT16 ro-enBLEU: 20.6198 -> 17.9401Hurt
raise Unigram bias strength delta from 0.5 to 2.0Unigram · WMT16 ro-enBLEU: 20.0175 -> 18.4838Hurt
raise KGW delta from 0.5 to 2.0KGW · open-ended detectionTPR@FPR=1%: 43.9% -> 96.57%Improved
raise Unigram delta from 0.5 to 2.0Unigram · open-ended detectionTPR@FPR=1%: 53.59% -> 97.85%Improved
use unbiased MCmark(n=10) instead of DiPmark(alpha=0.4)unbiased watermark · open-ended detectionTPR@FPR=1%: 74.11% -> 97.09%Improved
increase MCmark list length n from 10 to 100MCmark · open-ended detectionTPR@FPR=1%: 97.09% -> 92.2%Hurt
use DIPPER paraphrase attack instead of 10% random token replacementMCmark(n=10) · robustnessTPR@1%FPR: 96.11% -> 5.10%Hurt
use DIPPER paraphrase attackKGW(delta=2.0) · robustnessTPR@1%FPR: 91.68% -> 1.35%Hurt
use unbiased SynthID instead of Unigram(delta=2.0) for robustnessSynthID · robustnessTPR@1%FPR (random 30%): 95.63% -> 7.71%Hurt
use ITS-Edit instead of EXP-Editedit-distance unbiased watermark · WMT16 ro-enBLEU: 20.4879 -> 17.9904Hurt
raise Unigram delta from 0.5 to 2.0Unigram · unbiasedness, repeated prompts (SPMG)BLEU deviation: 0.8311 -> 3.9869Hurt
use SynthID instead of no watermarkSynthID · unbiasedness, repeated prompts (SPMG)BLEU deviation: 0.1199 -> 2.5266Hurt

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