To evaluate evolving knowledge injection in LMMs, we propose a pipeline to automatically collect evolving knowledge, constructing the EVOlving KnowledgE (EVOKE) benchmark. The EVOKE benchmark comprises 9,422 knowledge-image pairs for LMM knowledge injection, spanning 159 fine-grained types (29 New types and 130 Entity types), highlighting its diversity.
https://arxiv.org/abs/2505.24449
You can download EVOKE š¤. And the expected structure of files is:
EVOKE
|--ā¦ See the full description on the dataset page:
https://huggingface.co/datasets/kailinjiang/MMEVOKE.
