Generalist AI Releases GEN-1.5: A Robot Foundation Model That Learns New Tasks From One 3–12 Second Demo
One 12-second demo. That's all GEN-1.5 needs to teach a robot a new task—no fine-tuning, no retraining.
Why it matters
Robot foundation models are crossing a capability threshold: one-shot in-context learning from sensorimotor data is becoming real. This shifts the frontier of embodied AI from task-specific training to prompt-like adaptation—a model architecture breakthrough with immediate deployment implications.
The key facts
6 to knowGEN-1.5: robot foundation model learns new tasks from single 3-12 second video demonstration
One-shot in-context learning—no gradient updates or fine-tuning required
30-second context window for sensorimotor data
Tested across 10 diverse manipulation tasks
59% average success rate on one-shot trials
Released by Generalist AI
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: Generalist AI has released GEN-1.5, a robot foundation model that learns a new physical task from a single demonstration. Drop 3–12 seconds of sensorimotor data into its 30-second context window, and the robot performs the task. No gradient updates, no fine-tuning, no task-specific programming.…