Memory Less Agent
Memory-less agents in artificial intelligence research focus on developing systems capable of performing complex tasks without relying on persistent memory of past experiences. Current research explores efficient memory management strategies within agent architectures, such as incorporating workflow memories or employing task-based frameworks with StrictJSON for concise information handling, to improve performance on diverse tasks like web navigation and maze solving. These efforts aim to understand the fundamental limits of information processing without memory and to design agents that can effectively navigate complex, dynamic environments despite their memory constraints, advancing both theoretical understanding and practical applications in areas like robotics and natural language processing.
Papers
Asynchronous Agents with Perfect Recall: Model Reductions, Knowledge-Based Construction, and Model Checking for Coalitional Strategies
Dilian Gurov, Filip Jamroga, Wojciech Jamroga, Mateusz Kamiński, Damian Kurpiewski, Wojciech Penczek, Teofil Sidoruk
Unraveling the Complexity of Memory in RL Agents: an Approach for Classification and Evaluation
Egor Cherepanov, Nikita Kachaev, Artem Zholus, Alexey K. Kovalev, Aleksandr I. Panov