Improving Zero-shot Generalizationin Offline Reinforcement Learning using Generalized Similarity Functions