But these say exactly the opposite, the more tok/param the better. There is some optimum after which you need more training FLOPS to improve than if you add par

But these say exactly the opposite, the more tok/param the better. There is some optimum after which you need more training FLOPS to improve than if you add parameters but it is definitely not the other way around