During sleep and rest, the hippocampus replays some experiences and not others. Is it maximising future usefulness, or following context and structure?
Better replay rules could make AI learn more efficiently — and explain which memories you keep.
A reinforcement-learning rule reproduces many patterns of what gets replayed; newer work argues context and structure matter in ways the rule misses.
A neutral summary kept current by super users, not a verdict.
Replay prioritises what most improves future decisions ('gain × need').
Mattar & Daw 2018
Replay follows context and abstract structure, not reward value.
Zhou et al. 2025; Liu et al. 2019
No decisions yet. When this room changes a flow or an evidence label, it is recorded here permanently.
Dhi: Room opened.