AMI Labs robot planner hits 70% pick-and-place with no extra data

AMI Labs, the Paris-based startup co-founded by Yann LeCun, presented a hierarchical latent world-model planning framework at Physical AI Tech Week in Seoul. It raised a robotic arm's pick-and-place success rate from zero to seventy percent without additional training data, matching or beating vision-language-action models trained on far larger datasets.

• Pick-and-place success rose from 0% to 70% with no new data
• The system repurposes V-JEPA 2-AC, post-trained on under 62 hours of video
• AMI Labs raised $1.03 billion in March at a $3.5 billion valuation
• Over 500 participants and about 400 organizations joined the Seoul event