Jiahao Xu, Rui Hu, Olivera Kotevska, Zikai Zhang
View original ↗Build an open-source, multi-bit watermarking toolkit for LLM-generated text. This would allow developers to embed traceable metadata into outputs without degrading text quality, filling a massive gap for AI-tracing tools.
Suggested repo: xmark
"Watermark your LLM output without the gibberish."
Estimated effort: 50h