Agent Memory Architecture
Methods for designing, evolving, and repairing memory architectures for LLM agents based on execution traces and fine-grained feedback.
4 papers
Where this stands
The written synthesis of this thread is for subscribers. Subscribe.
How this thread developed
Oct 2026 · Stanford University, Carnegie Mellon University
LLM agents learn to organize document collections before questions arrive
Trains agents to pre-organize document collections into efficient hierarchies for proactive retrieval.
Oct 2026 · Alibaba Token Hub, Alibaba Group, The Hong Kong Polytechnic University
Co-evolved agents lose gains when their skill banks change
Examines impact of changing skill banks on agent performance, informing memory architecture design.
1 further paper
Oct 2026 · State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences
Execution traces enable precise localized repair of agent memory architectures
Introduces MemScope for evolving LLM agent memory architectures by inspecting execution traces.
3 of 4 papers shown