Misspelling "Native" Dynamic Resolution

[SiyangJ]

In the Qwen2-VL technical report, naive dynamic resolution references the paper Patch n’ Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution by Dehghani et al. However, NaViT stands for Native Resolution ViT according to the original paper. Therefore I believe "naive" dynamic resolution is an unintentional misspelling of "native" dynamic resolution.

PS: Not trying to nitpicking the paper. I've been fascinated by the model since the day it was released and have been using it on various tasks so far. It's arguably the best open source multimodal LLM as of today. Nice work. Just wanna leave a comment here in case other people might be confused by the naming like me.