AI BENCHMARK PROFILE

ReRef-3D

General AIMultimodal Perception

ReRef-3D benchmarks language-guided 3D scene rearrangement with 33,826 instructions across 998 scenes, evaluating placement validity and relation satisfaction.

Released: 2026-08-17
Readiness: Paper only
Primary field: General AI

Why it matters

It targets spatial reasoning in embodied AI and provides metrics for relation satisfaction and physical validity in rearrangement tasks.

Motivation

We introduce ReRef-3D, a benchmark for language-guided placement in 3D scenes.

Primary resources

Paper ↗Hugging Face ↗

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