Excuse me, where's the evaluation part?

[ZaiyiHu]

I failed to find the four mentioned (CF, IB, POS, ROT) score in the given code.

[Zjw-Louie]

I'm wondering if the author hasn't released the LLaVa pre-trained model, since I couldn't find any relevant links. Does this mean we can only run it using the original GPT-4o?

[ZaiyiHu]

I'm wondering if the author hasn't released the LLaVa pre-trained model, since I couldn't find any relevant links. Does this mean we can only run it using the original GPT-4o?

Probably, it seems many works are using GPT-4o API to do similar things. But I also think VLM itself can be replaceable, I'm currently using QWEN3-VL.