AI Paper Atlas

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

LLM Watermark Evasion via Bias Inversion

Jeongyeon Hwang, Sangdon Park, Jungseul Ok

ICML 2026 · Large language models › Language, safety and interpretability · approach: Text watermarking

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
apply negative logit bias beta=-4 to proxy suppression set during rewriting (ablation)LLM paraphrasing attack (Llama-3.1-8B) · C4, SIR watermark, Llama-3.1-8B-InstructASR: 54.2 -> 99.6Improved
strengthen negative bias beyond -4 (to -9) (ablation)BIRA · C4, SIR watermark, Llama-3.1-8B-InstructLLM judgment score: 4.24 -> 2.94Hurt
strengthen negative bias beyond -4 (ablation)BIRA · C4, SIR watermark, Llama-3.1-8B-InstructASR: 99.6 -> 99.6No clear effect
raise percentile q of suppression set (suppress fewer, higher-surprisal tokens) to 0.9 (ablation)BIRA · C4, SIR watermark, Llama-3.1-8B-InstructASR: 99.6 -> 77.2Hurt
raise percentile q to 0.9 (ablation)BIRA · C4, SIR watermark, Llama-3.1-8B-InstructLLM judgment score: 4.24 -> 4.60Improved
suppress all tokens (q=0) instead of q=0.5 (ablation)BIRA · C4, SIR watermark, Llama-3.1-8B-InstructLLM judgment score: 4.24 -> 4.15Hurt
select suppression tokens by self-information instead of random (ablation)BIRA · token selection, ratio 0.1-0.9ASRImproved
use BIRA instead of SIRA on 600-token textsrewriting attack · 600-token SIR text from Qwen2.5-32BLLM judge: 1.33 -> 4.14Improved
use BIRA instead of SIRA on 600-token textsrewriting attack · 600-token SIR text from Qwen2.5-32BTPR@FPR=1%: 20.8 -> 4.0Improved
use BIRA instead of vanilla paraphrasing (GPT-4o-mini)rewriting attack · C4, seven watermarksaverage ASR: 49.6 -> 99.5Improved
use BIRA instead of vanilla paraphrasingrewriting attack · SIR z-scorez-score: 0.19 -> -0.06Improved
shift SIR watermark signal to low-entropy tokens (ablation)SIR watermark defense · BIRA robustness, appendixdetectability-quality tradeoffHurt

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