Recent calls for pluralistic alignment emphasize that AI systems should address the diverse needs of all people. Yet, existing methods and evaluations often require sorting people into fixed buckets of pre-specified diversity-defining dimensions (e.g., demographics, personalities, communication styles), oversimplifying the rich spectrum of individualistic variations. To achieve an authentic representation of diversity that respects individuality, we propose individualistic alignment as a more tangible direction towards building AI for all by inferring individual preferences from the ground up.One prerequisite ability for approaching the individualistic alignment goal is to infer an individual’s general value and preference system by observing instances of their statements and behaviors. We introduce WorldValueGenome (ValueGenome), a dataset designed to evaluate language models (LMs) in reasoning about an individual’s value preferences in novel situations by learning from value-expressing statements from the same individual. ValueGenome transforms 253 unstructured survey questions from the influential World Value Survey (WVS) into a rich repository of 929 standardized natural language statements that capture the "human value genome" of 93K unique real humans worldwide. With the novel application of WVS with ValueGenome, our study exposes the critical gap of LMs in understanding and predicting individualistic human values, inspiring new arena of research challenges around individualistic value alignment that personalizes AI interactions towards individualistic preferences.