Research map / Large language models / Language, safety and interpretability
ARGH-Mark: Anchor-Synchronized Watermarking with Hamming Correction for Robust and Quality-Preserving LLM Attribution
He Li, Xiaojun Chen, Jingcheng He, Zhendong Zhao, Shuguang Yuan, Xin Zhao, Yunfei Yang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add anchor synchronization (ablation) | ARGH-Mark · C4, 200 tokens | match rate: -4.50 when removed | Improved |
| add RG-balanced vocabulary modulation (ablation) | ARGH-Mark · C4, 200 tokens | perplexity: +2.50 perplexity when removed | Improved |
| add Hamming error correction (ablation) | ARGH-Mark · C4, 200 tokens | robustness: -3.50 when removed | Improved |
| use ARGH-Mark instead of CyclicShift/ECC under token deletion | ARGH-Mark · C4 + OPT-1.3B | match rate: 78.0 -> 95.0 (vs CyclicShift) | Improved |
| use ARGH-Mark instead of CyclicShift/ECC for 32-bit messages on short texts | ARGH-Mark · C4 Guanaco-7B-HF / CNN-DailyMail, 32 bits, 200 tokens | match rate: CyclicShift 96.10 vs ARGH 96.00; 93.30 vs 90.00 | Mixed |
| use anchor-based linear detection instead of enumeration decoding | ARGH-Mark · Decoding time | seconds: 0.07 vs ECC 1.33 to 2.78 seconds | Improved |
| use longer text (384 vs 200 tokens) for 32-bit messages (ablation) | ARGH-Mark · Table 1 | match rate: 99.90 -> 96.00 (200 tokens, Guanaco) | Improved |
Closest papers by meaning
- Learning to Watermark: A Selective Watermarking Framework for Large Language Models via Multi-Objective Optimization (NeurIPS 2025 · similarity 0.88)
- Token-Specific Watermarking with Enhanced Detectability and Semantic Coherence for Large Language Models (ICML 2024 · similarity 0.88)
- BiMark: Unbiased Multilayer Watermarking for Large Language Models (ICML 2025 · similarity 0.88)
- PRO: Enabling Precise and Robust Text Watermark for Open-Source LLMs (NeurIPS 2026 · similarity 0.87)
- Theoretically Grounded Framework for LLM Watermarking: A Distribution-Adaptive Approach (NeurIPS 2025 · similarity 0.87)
- Provable Robust Watermarking for AI-Generated Text (ICLR 2024 · similarity 0.87)
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.
