VISTA-LLAMA: Reducing Hallucination in Video Language Models via Equal Distance to Visual Tokens | Read Paper on Bytez