all LLMs are undertrained to some degree. assuming the models are identical except one is bigger then the bigger model is better because 70% of a bigger number

all LLMs are undertrained to some degree. assuming the models are identical except one is bigger then the bigger model is better because 70% of a bigger number is larger than 70% of a smaller number. Now if you train a smaller model much longer than the bigger model (more tokens) then you are reducing the level of "under-trainedness" to some degree. at some point, you may have a smaller model that is better than that larger model. 70% of a bigger number may be larger than 70% of a smaller number but no guarantee 70% of a bigger number is larger than say 90% of a smaller number and so on.