reverse-SynthID-text

SynthID 水印逆向工程 概述 该目录包含用于分析和移除AI生成文本中SynthID水印的工具。该水印由Google DeepMind开发，并于2024年发表在《自然》杂志上。 SynthID 工作原理 水印生成过程 1. N-gram 上下文 ：对于每个词元位置，SynthID 将之前的 ngramlen - 1 个词元（默认：4个词元）视为上下文。 2. 哈希计算 ：计算以下内容的哈希： 上下文词元 候选的下一个词元 一组秘密水印密钥（默认30个） 3. G值分配 ：哈希用于为每个密钥层的每个可能的下一个词元分配一个二进制g值（0或1）。 4. 概率修改...

claude-awm

claude-awm: can you scrub a SynthID text watermark by editing the text? Yes, but only one family of attack works, and it isn't the one everyone assumes. Unicode variation selectors category Mn, U+FE00 to U+FE0F and U+E0100 to U+E01EF drive the detector below threshold and stay there. Every other...

HeavyWater and SimplexWater: Watermarking Low-Entropy Text Distributions

Large language model LLM watermarks enable authentication of text provenance, curb misuse of machine-generated text, and promote trust in AI systems. Current watermarks operate by changing the next-token predictions output by an LLM. The updated i.e., watermarked predictions depend on random side...

LLM-Text Watermarking Based on Lagrange Interpolation

The rapid advancement of LLMs Large Language Models has established them as a foundational technology for many AI and ML-powered human computer interactions. A critical challenge in this context is the attribution of LLM-generated text -- either to the specific language model that produced it or ...