In my opinion its a great idea because you get reuse hidden states which allows you to do something akin to thinking/[online learning]. If you only evolve use i

In my opinion its a great idea because you get reuse hidden states which allows you to do something akin to thinking/[online learning]. If you only evolve use input space outputs you're dealing with more decoding pressure. Also related to what I was saying see these: https://arxiv.org/pdf/2609.16372v1 https://arxiv.org/pdf/2609.16372v1 https://arxiv.org/pdf/2609.01449v1 https://arxiv.org/pdf/2609.01449v1 (this one is quite mindblowing cause they noise they the actual input space every time but it still works) https://arxiv.org/pdf/2609.11801 https://arxiv.org/pdf/2609.11801