New Research Advances Multi-Agent AI Systems with Skill Learning and Memory Integration

New Research Advances Multi-Agent AI Systems with Skill Learning and Memory Integration

Several new research papers are advancing multi-agent AI systems with sophisticated learning and memory capabilities.

According to arxiv.org, researchers have proposed EvoAgent, “an evolvable large language model (LLM) agent framework that integrates structured skill learning with a hierarchical sub-agent delegation mechanism.” The framework models skills as “multi-file structured capability units equipped with triggering mechanisms and evolutionary metadata,” and enables continuous skill generation through user feedback. The system incorporates “a three-stage skill matching strategy and a three-layer memory architecture” supporting dynamic task decomposition. In experimental results, “after integrating EvoAgent, GPT5.2 achieves significant improvements in professionalism, accuracy, and practical utility,” with the overall average score increasing “by approximately 28%” under a five-dimensional evaluation protocol.

In related work, arxiv.org describes MALMAS (Memory-Augmented LLM-based Multi-Agent System) for automated feature generation from tabular data. The system “decomposes the generation process into agents with distinct responsibilities,” with a Router Agent activating appropriate agent subsets. It integrates “a memory module comprising procedural memory, feedback memory, and conceptual memory, enabling iterative refinement.”

Additionally, arxiv.org presents FELA, “a multi-agent evolutionary system that autonomously extracts meaningful and high-performing features from complex industrial event log data.” FELA employs “specialized agents—Idea Agents, Code Agents, and Critic Agents—to collaboratively generate, validate, and implement novel feature ideas,” according to the paper.