SimSkill: A Self-Evolving LLM Agent for Skill and Knowledge Accumulation in Traffic Simulation
cs.AI, cs.MA
Submitted: 2026-09-03
Updated: 2026-09-11
Code: https://github.com/qiliuchn/SimSkill-V1
License: http://creativecommons.org/licenses/by/4.0/
The gist: As large language models (LLMs) become increasingly capable, the long-term value of AI systems depends not only on solving individual requests, but also on transforming experience and accumulated
Terminology
Abstract
As large language models (LLMs) become increasingly capable, the long-term value of AI systems depends not only on solving individual requests, but also on transforming experience and accumulated knowledge into durable, reusable competence. We introduce SimSkill, a self-evolving agent built around the Simulation of Urban MObility (SUMO) traffic simulator. SimSkill identifies capability gaps, generates and solves environment-grounded tasks, verifies solutions through an action--critic loop, and consolidates experience into episodic, procedural, and semantic memory without updating the backbone model. Through autonomous exploration, it builds a reusable library spanning the traffic-simulation workflow. We evaluate SimSkill on two held-out benchmarks with three backbone LLMs and independent artifact-based verification. SimSkill improves verified completion by up to 25 percentage points, while ablations show complementary contributions from procedural and semantic memory. Its benefits remain backbone- and budget-dependent: memory does not improve every model or uniformly reduce inference cost. More broadly, SimSkill illustrates a design paradigm in which natural language preserves and composes computational capabilities, while executable tools and code provide precise and reproducible execution. All code and experimental data are publicly available at https://github.com/qiliuchn/SimSkill-V1.
Sources
- Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory
- TrafficSimAgent: A Hierarchical Agent Framework for Autonomous Traffic Simulation with MCP Control
- Reinforced Self-Training (ReST) for Language Modeling
- AgentSUMO: An Agentic Framework for Interactive Simulation Scenario Generation in SUMO via Large Language Models
- MemGPT: Towards LLMs as Operating Systems
- A Survey on Self-Evolution of Large Language Models
- LLM-Assisted Light: Leveraging Large Language Model Capabilities for Human-Mimetic Traffic Signal Control in Complex Urban Environments
- Voyager: An Open-Ended Embodied Agent with Large Language Models
- SkillOpt: Executive Strategy for Self-Evolving Agent Skills
- AutoSkill: Experience-Driven Lifelong Learning via Skill Self-Evolution
- SUMO-MCP: Leveraging the Model Context Protocol for Autonomous Traffic Simulation and Optimization
- ReAct: Synergizing Reasoning and Acting in Language Models
- LifelongAgentBench: Evaluating LLM Agents as Lifelong Learners
- Ghost in the Minecraft: Generally Capable Agents for Open-World Environments via Large Language Models with Text-based Knowledge and Memory
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
- VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection