Found 1 document(s)
Page 1 of 1
-
Long-horizon LLM agents face a fundamental allocation problem as interaction histories grow: which information to retain, how much context to devote to it, and at what granularity. Existing memory mechanisms largely focu...