COUNTERMEM: World-Model Verified Counter-Factual Memory for Language Agents
C COUNTERMEM is introduced, a reinforcement-learning framework for constructing and using verified counterfactual memory across tasks, and it is shown that removing verification or persistent storage weakens the gains, while applying verified corrections to unsuitable decisions can reverse them.
Hong-Ji Pu, Rui-Xiang Tang, Yong-Feng Zhang
· 0 citations