Aurora-M

AI & ML interests

We present a set of red-teamed models. Trained on the LUMI HPC in Finland (thus the name Aurora). The "-m" designation stands for multimodal, multilingual, multidomain mixture of expert (MOE) models, each of which we intend to research. As part of Ontocord.AI's dedication to lawful open science AI, we coordinated this effort of volunteers and contributed to the safety measures. This work should NOT be confused with the AuroraGPT, https://www.hpcwire.com/2023/11/13/training-of-1-trillion-parameter-scientific-ai-begins/.

Recent Activity

felfri  authored a paper 2 days ago

LLMs Lost in Translation: M-ALERT uncovers Cross-Linguistic Safety Gaps

PSaiml  authored a paper 2 days ago

LLMs Lost in Translation: M-ALERT uncovers Cross-Linguistic Safety Gaps

sted97  authored a paper 2 days ago

What's the Meaning of Superhuman Performance in Today's NLU?

View all activity

Team members 40