Sunday Robotics Unveils ACT-2 Home Robot Preview
tonyzzhao · x · 2026-07-18
Sunday Robotics released ACT-2 Preview, emphasizing it as the first robotic model to balance broad generalization with high reliability in real homes.
Key points include:
- Fine-tuning with just one example is enough to generalize a new behavior to unseen environments.
- Introduction of the new "Solve" framework: measuring progress via "declared scope + adaptation cost"; for new homes, the adaptation cost is zero.
- Scaled training significantly narrows the gap between in-lab performance and real-home performance, indicating that larger pre-training better predicts real-world results.
- Through parallel iteration across multiple robots, the team can scale performance/quality/speed faster and more reliably.
The first task is laundry folding:
- Achieved a 99.1% success rate across 31 unseen homes and 785 trials.
- Average quality of 4.72/5, approaching human level.
- Also demonstrated robustness in floor pickup, varying sizes, distractions, and lighting changes.
- The same model has already started learning subsequent skills like vacuuming and toy organizing.
Related event: Sunday Robotics Unveils ACT-2 Preview for Home Robots(17 posts)→
More from Embodied
- Hugging Face Teases Agentic Training Environments with OpenEnv for August Launch — mervenoyann · 2026-07-22
- NVIDIA shows 22 SIGGRAPH papers and Omniverse tools for robot simulation — facontidavide · 2026-07-22
- Nothing phone mockup turns a film joke into a modular design meme — ZeYanjie · 2026-07-22
- Lightwheel AI Launches SimReadyGen: Text-to-Physics-Accurate Robot Sim Assets — ZeYanjie · 2026-07-22
- Humanoid robot sorting packages in a warehouse sparks debate over job loss — MonaJalal_ · 2026-07-22
- NVIDIA pushes OpenUSD as the common layer for simulation and physical AI — MonaJalal_ · 2026-07-22