Reproduction: From Prompts To Tokens: Internalizing Causal Supervision In Vision Language Model For Multi Image Causal Reasoning | Genaihub