Benchmark RadarAI BENCHMARK PROFILE
MMHBench
MMHBench is a multimodal benchmark for mental health understanding in long-form videos, comprising 268 videos and 2,184 questions across third-person and first-person settings. It evaluates model reasoning about mental states from multimodal evidence.
- Released
- 2026-07-30
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
This benchmark targets a critical domain where existing evaluation is limited to coarse classification. It could provide finer-grained assessment of mental health understanding, but its current availability is unclear.
Motivation
Mental health understanding in long-form videos requires nuanced reasoning over observable behavior, interpersonal context, and latent psychological states.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.