DeepMind reveals Gemini Robotics 2.5 and open safety benchmark "ASIMOV-jailbreak"
Ars Technica

DeepMind has announced Gemini Robotics 2.5 and an open safety benchmark called ASIMOV-jailbreak.
Google's DeepMind has revealed the new Gemini Robotics 2.5 model, which achieves up to twice the success rate on certain tasks compared to previous versions. Alongside the model, the organization introduced an open safety benchmark named ASIMOV-jailbreak. The developments aim to advance robotic capabilities while addressing safety concerns.
Why it matters: The release of Gemini Robotics 2.5 and the ASIMOV-jailbreak benchmark represents ongoing advancements in robotics performance and safety evaluation.




