What Must Generalist Agents Remember?

In three linesTheoretical paper on memory requirements for generalist agents. Proves that agents performing near-optimally across multiple domains must maintain distinct memory distributions at observational bottlenecks. Memory enables domain disambiguation, transition-model reconstruction, and planning.

Summary generated by Claude — human-verified