Benchmark RadarActiveFly-Bench
ActiveFly-Bench is a benchmark for UAV embodied perception, decomposing active perception into three tasks: Aerial Embodied Question Answering (Air-EQA), Observation Behavior Planning (OBP), and Fine-grained Language-guided UAV Control (FLUC). It includes datasets from real-world and simulated outdoor environments.
- Released
- 2026-07-11
- Readiness
- Paper only
- Primary field
- Robotics & Autonomous Systems
Why it matters
Active perception in UAVs requires bridging high-level reasoning with low-level control, which current VLMs and VLA models struggle with. This benchmark provides a testbed for embodied aerial intelligence, addressing the lack of benchmarks in this domain.
Motivation
We introduce ActiveFly-Bench, the first benchmark to bridge cyberspace reasoning and physical-world interaction for UAV embodied perception.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.