VVM-Bench
VVM-Bench evaluates Large Multimodal Models on semantic perception and modality understanding across six real and synthetic modalities, using multiple-choice questions and generation tasks to assess zero-shot generalization to unseen visual modalities.
- Released
- 2026-07-11
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
VVM-Bench provides a standardized protocol for assessing LMMs' ability to generalize across visual modalities, enabling comparison of current models on a crucial capability for real-world deployment where sensor types vary.
Motivation
Despite the advancements of Large Multimodal Models (LMMs) in RGB vision, their ability to generalize to unseen visual modalities remains a largely unexplored challenge.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.