neulab/Mind2Web_train_llava
Mind2Web training set for the paper: Harnessing Webpage Uis For Text Rich Visual Understanding š Homepage | š GitHub | š arXiv Introduction We introduce MultiUI, a dataset containing 7.3 million samples from 1 million websites, covering diverse multi- modal tasks and UI layouts. Models trained on MultiUI not only excel in web UI tasksāachieving up to a 48% improvement on VisualWebBench and a 19.1% boost in action accuracy on a web agent dataset Mind2Webābutā¦ See the full description on the dataset page: https://huggingface.co/datasets/neulab/Mind2Web_train_llava.
Conversations for this repository live on Hugging Face.
CoolFace shows imported repositories read-only. Posting into someone elseās repository from here would need an authorised integration and the account holderās consent, so the link goes to the source instead.
Open discussions on Hugging Face