Benchmark RadarAI BENCHMARK PROFILE
InstructMove
Evaluates instruction-following in robot manipulation through pick-and-place tasks requiring language grounding across category, attribute, spatial, and compositional skills.
- Released
- 2026-08-24
- Readiness
- Runnable
- Primary field
- Robotics & Autonomous Systems
Why it matters
Provides a controlled testbed to diagnose visual shortcut learning in vision-language-action policies, improving reliability of language-conditioned manipulation.
Motivation
Vision-language-action (VLA) models have made general-purpose robot manipulation increasingly plausible by conditioning robot actions on natural-language instructions.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.