FuxiaoLiu / MMC

[NAACL 2024] MMC: Advancing Multimodal Chart Understanding with LLM Instruction Tuning
82 stars 3 forks source link

question about MMC Dataset #19

Closed hkunzhe closed 2 months ago

hkunzhe commented 2 months ago

As mentioned by #9 , the updated MMC-Benchmark in huggingface still only contains true-or-false questions. Since the original issue was closed, I reopen this one.

FuxiaoLiu commented 2 months ago

You can check the data right here link.

hkunzhe commented 2 months ago

You can check the data right here link.

@FuxiaoLiu Thanks for your reply! By the way, I found there are some duplicated question in https://huggingface.co/datasets/xywang1/MMC/viewer/MMC-Benchmark/

image

Is this normal?

FuxiaoLiu commented 2 months ago

You can check the data right here link.

@FuxiaoLiu Thanks for your reply! By the way, I found there are some duplicated question in https://huggingface.co/datasets/xywang1/MMC/viewer/MMC-Benchmark/ image Is this normal?

You can skip the duplicated question if the image is the same. I use "random select" method to select the topic for the topic classification tasks. Therefore, as for some images, the questions might be the same.