Large Universe Model/Comparisons
Large Universe Model comparisons
The Large Universe Model is defined by what it does that the Large Language Model and the Large World Model cannot. These pages set it against each predecessor and each near-neighbour in turn.
Large Universe Model vs Large World ModelThe Large World Model simulates a scene it is shown. The Large Universe Model ingests continuously and revises its beliefs. A full comparison of LUM vLarge Universe Model vs Large Language ModelThe Large Language Model is trained once and frozen at a cutoff. The Large Universe Model ingests continuously and has no cutoff. A full comparison ofLarge Universe Model vs retrieval-augmented generationRAG fetches documents at query time to patch a Large Language Model’s training cutoff. A Large Universe Model never had a cutoff. Why retrieval and coLarge Universe Model vs fine-tuningFine-tuning folds new material into a Large Language Model in batches, on a schedule. A Large Universe Model revises on every observation. Why periodiLarge Universe Model vs online and streaming machine learningOnline learning and streaming ML have updated models continuously for decades. A Large Universe Model generalises the pattern: the same loop without aLarge Universe Model vs AI agentsAn agent executes a task and terminates. A Large Universe Model attends continuously and holds beliefs between tasks. Why agents and Large Universe Mo