Research · memo · added 9 Mar 2026
Trajectory-Informed Memory Generation for Self-Improving Agent Systems
A research note, not a framework evaluation. It carries no architecture score and is not part of the directory or the comparison table.
Agents repeat mistakes. They fail the same way, miss optimization opportunities, and don't transfer successful strategies across tasks. This paper proposes a 4-component pipeline that automatically extracts structured, typed learnings from execution trajectories and injects them into future agent prompts via similarity-based retrieval.