Skild AI publishes S1 in-context robot learning model
Skild AI published S1, a robotic foundation model that learns new tasks from a single visual demonstration without fine-tuning or post-training.
Evidence notes
- S1 applies in-context learning to seen and unseen manipulation tasks, including plant potting, pancake preparation, pour-over coffee and kit assembly.
- Skild reports unseen tasks lasting up to 10 minutes and a sevenfold improvement over language-only prompting on its evaluation set.
- The official release demonstrates transfer from a human video example to robot execution, including an 11-minute demonstration-to-execution plant-potting workflow.
Company context
Skild AI develops foundation models and AI systems designed to enable general-purpose robot intelligence across different hardware platforms.