Robotics Foundation Model Progress: Skild S1 Enables In-Context Learning
deepakpathak · x · 2026-08-27
Charts from Skild, Generalist, and Sunday published over the last five weeks finally allow for objective measurement of progress in robotics foundation models.
- Sunday (ACT-2): Achieves 99.1% success on folding laundry in unseen homes. However, this is a single task chosen by the lab, and the curve is hitting a ceiling, suggesting scaling only perfects a specific skill.
- Generalist (GEN-1.5): Shows loss falling over 8 months of training and 59% one-shot success on 10 novel tasks. While the loss chart tracks prediction error, it does not directly equate to task success, and 59% is a single data point that doesn't confirm scaling laws.
- Skild (S1): Introduces a new foundation model that learns via in-context learning. It can be taught 10-minute long tasks it has never seen before from a single video prompt without any fine-tuning, operating in real-time.
More from Embodied
- Viral meme nails AI's irony: it can write symphonies but can't do the dishes — briannekimmel · 2026-10-16
- WEDGE turns navigation agents' own failures into targeted training worlds — mohitban47 · 2026-10-16
- Knödel Watch: animated vocabulary cards on ESP32-S3, open-sourced 4 days after unboxing — alexcovo_eth · 2026-10-16
- SlipSense fuses pressure and vibration for 23ms zero-shot robotic slip detection — micoolcho · 2026-10-16
- Stanford's Fei-Fei Li team open-sources OpenWAM, a unified world-action model framework for robots — jiqizhixin · 2026-10-16
- Robot folding leaderboard: GPT-6 Astra leads at 1349 Elo, Muse Wonder trails at 1034 — qinzytech · 2026-10-16