Benchmark Radar
AI BENCHMARK PROFILE

ContactWorld

Robotics & Autonomous SystemsRobotics & Embodied Intelligence

ContactWorld is a benchmark and empirical study for vision-tactile world models in contact-rich manipulation, spanning 12 tasks including insertion and screwing, evaluating planning success rates.

Released
2026-06-11
Readiness
Paper only
Primary field
Robotics & Autonomous Systems

Why it matters

Addresses the gap in understanding which representation properties support stable long-horizon planning in contact-rich settings, offering practical guidance for multimodal model design.

Motivation

Contact-rich manipulation requires world models to reason over complex contact dynamics from multimodal sensory observations.

Primary resources

Benchmark Radar records only publicly supported details and links back to primary sources for verification.