Benchmark RadarPOINav-Bench
POINav-Bench evaluates vision-language navigation agents in real-world POI-goal navigation across 11 reconstructed commercial areas covering 126,398 m² with 163 POIs, using traversability-aware annotations and reference trajectories for closed-loop evaluation.
- Released
- 2026-05-27
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Existing VLN benchmarks for POI-goal navigation suffer from coarse granularity or sim-to-real gaps. POINav-Bench provides high-fidelity real-world environments, enabling evaluation of final-meters navigation capabilities that are critical for practical deployment.
Motivation
Real-world navigation is fundamentally driven by Points of Interest (POIs), yet reaching a precise POI remains a critical "final-meters" challenge.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.