Introducing S1: In-Context Learning for Robotics | Skild AI
S1 is our flagship robotic foundation model, built from the ground up as an in-context learner: show it a single video demonstration of a task, seen or unseen, short or long-horizon, and it executes with no fine-tuning or post-training.
0:00 / 0:00 Unseen tasks10-minute horizonsOne video promptNo post-training 13-minute read Introduction The evolution of language modeling provides a blueprint for turning capable models into useful general-purpose tools. Earlier approaches based on transformers, such as BERT (Devlin et al., 2019), proved effective at understanding language, but each new application still demanded additional data collection and fine-tuning of the model. The pivotal transition from such early language models to ChatGPT was driven by the emergence of a fundamentally different learning paradigm, a phenomenon…
saved by
related reading
- [2607.15275] RoboTTT: Context Scaling for Robot Policiesarxiv.org
- GEN-1.5: Embodied Foundation Models are One-Shot Learners - Generalist AIgeneralistai.com
- State of Robot Learning, December 2025vedder.io
- ACT-1: A Robot Foundation Model Trained on Zero Robot Data | Sunday Robotics | The helpful robotics companysunday.ai
- How Claude Performs on Robotics Tasks \ Anthropicanthropic.com
- A Steerable Model with Emergent Capabilitiespi.website
- Generalist - GEN-0 / Embodied Foundation Models That Scale with Physical Interactiongeneralistai.com
- Emergence of Human to Robot Transfer in Vision-Language-Action Modelspi.website
- How does in-context learning work? A framework for understanding the differences from traditional supervised learning | SAIL Blogai.stanford.edu
- RoboTTT: Context Scaling for Robot Policiesresearch.nvidia.com
- GitHub - robotics-survey/Awesome-Robotics-Foundation-Modelsgithub.com
- Causal Video Models Are Data-Efficient Robot Policy Learners | Rhoda AIrhoda.ai