- KRONOS-1M keeps one unbroken session across terrain, 60 Hz motion, LiDAR, and operator rubrics inside a 1,000,000-token context. - It splits continuous robot trajectories into timed sub-tasks and writes past-and-future action captions for each segment. - It scores the whole episode, then labels every 0.5 second step as normal, suboptimal, or failure. - It grades trajectories against natural-language rubrics, and raises its thinking level when a step fails. - On a failure it checks physics in a sandbox and writes four counterfactual rollouts (one chosen, three rejected), including physics and robot-shape changes, as preference-training pairs.
More from Gemini 4 Argon SF Hackathon
TRY THE DEMO LIVE! labtwin.vercel.app note: the hosted demo replays a cached run. non-demo data requires requires a local server and API key: please see the…
Morf helps students, makers, and robotics teams turn ideas into interactive 3D assemblies through conversation. Describe an object, explore its individual…
Eagle makes flying a small plane as easy as getting in a car. Every hour in the air takes about five hours of planning on the ground. Eagle does that planning…
BEAMTEAM is a self-driving optical engineering system that turns a published figure into a buildable, verified laser-optics bench in under three minutes…
Full game: https://skywake.vercel.app/ Game studio workflow: https://skywake.vercel.app/studio Built a full 3D Game using Gemini Argon and created a fully…
Revive brings back video games that can no longer be played, using only the gameplay footage that survives on YouTube. Gemini 4 Argon finds a lost game's…
