Benchmark RadarAI BENCHMARK PROFILE
NextMotionQA
NextMotionQA is a benchmark for human motion understanding with VLMs, featuring three tasks: multiple-choice QA, video captioning, and fine-grained error correction. It includes expert-verified annotations across three semantic axes and three complexity levels.
- Released
- 2026-06-03
- Readiness
- Paper only
- Primary field
- Robotics & Autonomous Systems
Why it matters
Existing motion benchmarks have coarse granularity and ambiguity. NextMotionQA provides structured tasks across complexity levels, enabling diagnosis of VLM capability gaps in motion understanding and judging.
Motivation
Reliable evaluation of human motion understanding is fundamental to advancing embodied AI, robotics, and animation.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.