Self-Evolving Software Agents

2026-04-29

AI Summary

This work addresses the limited evolvability of traditional autonomous agents, whose requirements, goals, and capabilities are typically fixed at design time. To overcome this constraint, the authors propose a novel hybrid BDI-LLM architecture that deeply integrates Belief-Desire-Intention (BDI) reasoning with large language models (LLMs). This integration enables agents to autonomously extract new requirements from experience, generate corresponding goals, and synthesize executable code during runtime. By incorporating an automated evolution module within a continuous reasoning loop, the prototype system demonstrates successful behavioral self-evolution in dynamic multi-agent environments. The results validate the feasibility of LLM-driven agent evolution while also highlighting current limitations concerning behavioral inheritance and stability.

Abstract

Autonomous agents can adapt their behaviour to changing environments, but remain bound to requirements, goals, and capabilities fixed at design time, preventing genuine software evolution. This paper introduces self-evolving software agents, combining BDI reasoning with LLMs to enable autonomous evolution of goals, reasoning, and executable code. We propose a BDI-LLM architecture in which an automated evolution module operates alongside the agent's reasoning loop, eliciting new requirements from experience and synthesizing corresponding design and code updates. A prototype evaluated in a dynamic multi-agent environment shows that agents can autonomously discover new goals and generate executable behaviours from minimal prior knowledge. The results indicate both the feasibility and current limits of LLM-driven evolution, particularly in terms of behavioural inheritance and stability.

Problem

• self-evolving agents
• autonomous evolution
• software evolution
• goal adaptation
• BDI agents

Innovation

• self-evolving agents
• BDI-LLM architecture
• autonomous evolution
• goal discovery
• executable code synthesis