Unsupervised value iteration — is mirrored in model based form by → Unsupervised model learning

Unsupervised value iteration and unsupervised model learning are the same idea — extract information from the transitions of unlabeled episodes even though their rewards are unknown — applied to reinforcement learning's two dominant paradigms. The value-iteration version presupposes a model-free agent and uses unlabeled transitions to sharpen its Bellman updates directly; the model-learning version presupposes a model-based agent and uses the identical unlabeled transitions to improve the learned dynamics model it plans through (concrete-problems, §"5 Scalable Oversight", p. 12). Which of the two a builder reaches for is therefore dictated entirely by whether their base RL agent is model-free or model-based, not by anything specific to the semi-supervised oversight setting itself; the two are companions covering the same design space rather than rival techniques.