Currently, VLM is an area that attracts the attention of many research groups around the world. However, to carry out research projects, it requires a large amount of resources to train effective models. Realizing this problem, our group proposed an optimal sample filtering method, which has proven to be more effective than other methods, to eliminate noisy samples for VLM. This method helps to reduce training time, especially suitable for those with… See the full description on the dataset page:
https://huggingface.co/datasets/Nhanvi282/HQD4VLM.
