Efficient Inference
去掉 ViT，真的会更强吗？
一轮现有开源 ViT-free / encoder-free VLM 的统一横评：从模型能力、输入范式到视觉信息如何被任务真正利用。
#ViT-free#VLM#Video Understanding#Efficient Inference
A FIELD NOTEBOOK
Blog.
Research notes, technical readings,
and the practice around them.
All postsAll postsLong Video UnderstandingLong Video UnderstandingEfficient InferenceEfficient InferenceAgent SystemsAgent SystemsExperience SharingExperience Sharing