AI brief
A crowdsourced game built on Olmo 3 let players find unexpected model behaviors, stress-testing prosocial AI evaluations. The work indicates that open access to a model's internals can help researchers understand why those tests break.
Written by AI from Ai2 Blog's published text.
Source
“A crowdsourced game built on Olmo 3 showed how people can exploit unexpected model behaviors to stress-test prosocial AI evaluations—and how open access to a model’s internals can help researchers understand why those tests break.”
A paper, laboratory publication or research report.
Read original story ↗