Three seconds to learn: Gen1.5's few-shot skill learning may hinge on one attention masking rule
KyeGomezB · x · 2026-08-20
GEN is the first model line where one-shot and few-shot learning of physical skills appears to emerge from pretraining alone — three seconds of demonstration and it performs the task with no gradient updates. The author hypothesizes the key is a simple masking rule: an action token at time t may only attend to sensor tokens at t−δ, where δ is the inference latency sampled during training. This could explain how GEN avoids a System 1/System 2 split and inference-time guidance while hitting 100 Hz with a 7B model.
More from Embodied
- LLM-Controlled Robots Have <10% Chance of Reliability by 2028 — a_karvonen · 2026-08-20
- Using RealSense 3D camera for virtual green screen with depth control — chrismatthieu · 2026-08-20
- GEN-1.5 success driven by repetitive motion data and UMI collection — DrJimFan · 2026-08-20
- NVIDIA Tutorial: Post-Train Cosmos 3 Edge for On-Device Robot Control on Jetson Thor — MonaJalal_ · 2026-08-20
- Open-Gen Attempts to Replicate GEN 1.5 Embodied Model Architecture — KyeGomezB · 2026-08-20
- WRC 2026 pivots to real work: UBTech humanoids clock in 9-to-5 on the show floor — 智东西 · 2026-08-20