Back to home

Tag

sim-to-real

Sim-to-real focuses on transferring policies or models from simulation to real-world use while reducing the gap in dynamics, sensing, and control. It shows up in domain randomization, digital twins, synthetic trajectory generation, robot manipulation, and physics-based reasoning.

2 articles

Physics Simulators as RL Data for LLM Reasoning
Research/Apr 14

Physics Simulators as RL Data for LLM Reasoning

Researchers train LLMs on synthetic physics from simulators and report zero-shot gains on IPhO problems, showing a new path beyond web QA data.

SIM1 turns sparse demos into deformable-world data
Research/Apr 10

SIM1 turns sparse demos into deformable-world data

SIM1 grounds deformable-object simulation in real scenes, then scales sparse demos into synthetic training data for data-efficient robot policy learning.