EMPIRIC Agent Learns Residual World Models to Fill Physics Engine Gaps

Researchers from Basis Research Institute, Cambridge, Princeton and MIT introduced EMPIRIC, an agent that learns residual world models to capture mechanisms missing from physics engines, such as glue curing and heat effects. It passed all 25 levels across 5 simulation domains.

2026-10-02 ~ 2026-10-04 · 2 related posts

• EMPIRIC teaches robots missing physics as code, solving all 25 tasks where baselines get 14-16 — tomssilver · 2026-10-02

1 near-duplicate retellings: tomssilver