Long-Task Agents

Let AI Decide What to Remember — MemPO Uses RL to Cure Agent Amnesia

MemPO's diagnosis: the problem isn't capacity—nobody taught the model what's worth remembering. RL fixes this by letting models manage notes themselve