LocalLLaMA

Previously I made a post about my new model, IS-LM 3B. If you want more information of the model, read the post here.

As the creator of this model, I noticed that this model is extremely good for a 3B in economic tasks(As it is trained on it.), and surprising other tasks too. It follows instructions well, is VERY verbose, and is very accurate compared to other 3B models and even LLaMA 1 7B. Though it is bad at casual chats/roleplay, but that was expected because it was not trained on those. I am very surprised and impressed, and here's some example conversations(Warning: a LOT of text.): https://pastebin.com/2BZC7kZg

I did NOT cherrypick those, these are almost all of the proper tests I've done to the model. It is very consistent at generating those type of responses, if you don't believe me, try it out yourself. Also note that all these are seperate conversations, as I found that this model isn't the best at multi-turn conversations.

Obviously not all of them are correct or the best, but it does show very good capabilities and verbosity. This model is again, extremely good for a 3B, even outperforming a lot of the SOTA LLaMA 1 7B models IMO. If you are interested, I HIGHLY recommend you trying it out as it can be very easily ran.