Fetching the paper…

BenchLMM: Benchmarking Cross-style Visual Capability of Large Multimodal Models · Around