aiqbench analysis
SimSkill represents a significant advancement in autonomous agent development by enabling LLMs to self-evolve within traffic simulation environments without modifying their core weights. By consolidating experience into distinct memory types, the system demonstrates how durable competence can be built through iterative task generation and verification rather than continuous model retraining. This approach offers a practical pathway for scaling AI agents to complex, domain-specific workflows like urban mobility planning.
SimSkill: A Lifelong Learning AI Agent for Autonomous Mastery of Traffic Simulation
arXiv:2609.03753v1 Announce Type: new 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.
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 this https URL.
Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
arXiv:2609.03753 [cs.AI]
(or arXiv:2609.03753v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2609.03753 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 277 words
Related coverage
AutoWorld: Learning Multi-Agent Traffic Simulation with Self-Supervised World Models
arXiv · Jul 28, 2026
NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI
NVIDIA · May 21, 2026
GNN-based Multi-Agent Control of Traffic Shockwaves in Sparse Vehicular Ad-hoc Networks
arXiv · Jul 28, 2026
TerraZero: Procedural Driving Simulation for Zero-Demonstration Self-Play at Scale
arXiv · Aug 6, 2026