Directory
Published directory entries matching your search.
Tensor Trust is the source code for a web game and data cleaning pipeline described as a prompt injection attack game that collects data for adversarial ML research. The repository covers the game and pipeline and links to a paper and the game itself.
Tensor Trust is a prompt injection attack and defense game that invites users to fool AI language models and climb a leaderboard. The site is part of a research project studying prompt injection vulnerabilities in AI systems, using crowdsourced submissions to help researchers build more reliable AI.
