Event
Authors release JRDB-AVR active visual reasoning benchmark
1 reports · 1 sources
Zhixi Cai and other authors propose and release the JRDB-AVR active visual reasoning benchmark, along with the accompanying JRDB-AVR-Agent reference method. The benchmark is built on existing real JRDB robot data through a structured question generation engine. In the evaluation, embodied agents receive visual reasoning questions and must request limited observations by timestamp and viewpoint; the assessment examines both the final answer and the visual evidence supporting the answer. The authors state that the code and benchmark have been made public.
Report timeline
- JRDB-AVR: An Active Visual Reasoning Benchmark for Embodied Agents in Real-World Environments
The authors propose JRDB-AVR, an active visual reasoning benchmark built on real JRDB robotics data: after receiving a visual reasoning question, an embodied agent must request limited observations by timestamp and viewing angle, and evaluation examines both the final answer and the visual evidence supporting it.
arXiv cs.RO ·