Sunday previews ACT-2 robot foundation model
Sunday Robotics previewed ACT-2, a robot foundation model that combines broad generalization with iterative post-training, reporting 99.1% zero-shot success on laundry folding across diverse unseen homes.
Why it matters
ACT-2 targets the reliability gap between impressive robot demonstrations and repeatable household deployment by making small post-training improvements transfer across unseen homes.
Evidence notes
- Sunday said a single fine-tuning demonstration can teach ACT-2 a new behaviour that transfers to held-out garments and unseen environments.
- ACT-2 uses a pretrained model built from Sunday's sensorised human-motion dataset, then a rapid post-training loop to improve difficult edge cases on Memo robots.
- The company is taking ACT-2 into real homes through its Memo beta programme while extending the shared model to vacuuming, toy organisation, zippers, garment handling and coffee preparation.
Company context
Sunday Robotics is building general-purpose robots designed to operate in home environments, combining hardware and control systems into a single platform. A core part of its approach is collecting real-world human data using glove-based systems, capturing how tasks are performed and using that data to train robot behaviour.