Hi
I was working on POC for data analytics where Pandas AI is used as interpreter and get python code from LLM.
It worked well for basic use case like calculation , average , percentage but not giving good result if ask trends , summary of data , graph
Is there any other solution / liberary which can provide better data analytics result from LLM.
Just posted about this, I didn’t see your post. I agree this is highly relevant. I imagine a new architecture or finetune would be great for this purpose.
I believe what amitgoyal14 is suggesting is a model which performs inference on the data directly.
thanks . new architecture required new code interpreter. is there any better code interpreter ?
It would be interesting to have a model whose architecture is trained on a symbolic mathematic formula system, instead of English, as an intermediary to a normal LLM. I don’t know how such a system would be implemented, but I would be curious to see if there could be a model which is trained to take large datasets, form relationships, define formulas, etc. which are then used to produce an English output.
Hopefully they explore this idea!