penghao-wu / vstar

PyTorch Implementation of "V* : Guided Visual Search as a Core Mechanism in Multimodal LLMs"
https://vstar-seal.github.io/
MIT License
497 stars 32 forks source link

Prompt for GPT4V #11

Closed LengSicong closed 2 months ago

LengSicong commented 6 months ago

Hi authors, congrats on this great work!

May I know what your prompt is for evaluating GPT4V? We tested ourselves but found that the results were pretty different, especially the spatial relationship subset (where the accuracy is even far less than 50% for two-option MCQs).

penghao-wu commented 6 months ago

Hi, the prompt we used for evaluating GPT-4V is: You will be given a question and several answer options. You should choose the correct option based on the image provided to you. You just need to answer the question and do not need any information about individuals. When you are not sure about the answer, just guess the most likely one. Question: XXX Options: A. XXX B. XXX