search

BAAI-DCAI / Bunny

A family of lightweight multimodal models.
Apache License 2.0
799 stars 61 forks source link

模型无法理解用户输入了多少张图片 #56

Closed CanvaChen closed 2 months ago

CanvaChen commented 2 months ago

image 如图，我选了一张图片，回答说3张。

RussRobin commented 2 months ago

Hi @CanvaChen , thank you for trying our model out and this interesting failure case.

Currently in our training set, such QAs are not included. Our pretrain and fine-tune dataset include no more than 1 image as input. Our questions didn't ask the model to count number of input images. Again, this is a super interesting failure case, and thank you very much for sharing it!

Regards Russell BAAI