Questions about the paper

[ResonWang]

Dear authors,
I have some questions about the paper content:
(1) what is the MedVicuna and RadVicuna in Table 1? I cannot find them in the paper or on the Internet;
(2) According to Figure 1, it seems only the Linear Transformation Layer is trained in the whole framework, but why you mentioned in the contributions that "The LLM (Vicuna) is fine-turned on medical data"?
(3) In your work, if only the Linear Transformation Layer is trained while the LLM and MedClip are all frozen?

[OmkarThawakar]

Dear @ResonWang

Thanks for interest in our work.

Our MedVicuna and RadVicuna is fine-tuned version of vicuna on medical and radiologist conversation data provided.
We fine-tuned llm (vicuna) on medical and radiology conversation data separately prior to the training of linear layer.
We kept Llm and Image Encoder frozen while training linear layer.

[drahmedawais]

Hi Dear Authors,

So whether you trained LLM or Image encoder in this training stage-1 or stage-2?

[OmkarThawakar]

Dear @awaisahmednuces,
We train llm separately on medical and radiology conversation.
In our method, in stage-1 and stage-2 training, we train projection layer between image encoder and llm on MIMIC & OpenI data.