MM-Vet v2: A Challenging Benchmark to Evaluate Large Multimodal Models for Integrated Capabilities | Synapse