In this paper, we initially discuss about the ValueEval task and the challenges involved in multi-label classification tasks. We tried to approach this task using Natural Language Inference and proposed a Grouped-BERT architecture which leverages commonality between the classes for a multi-label classification tasks.