See2Think: Do Multimodal Models Really Use Intermediate Visual States? | Digital Library | PAMCET | PAMCET