“SmolVLA” A vision-language-action model for affordable and efficient robotics 2025年09月25日 Mishima Marina dls-2025, papers 【DL輪読会】“SmolVLA” A vision-language-action model for affordable and efficient robotics by @DeepLearning2023