# Code World Models for Parameter Control in Evolutionary Algorithms

arXiv: https://arxiv.org/abs/2602.22260 · HTML: https://mlsanity.com/paper/2602.22260/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in World models for language agents, one of 11 directions in World models.
- Builds on 9 of its 24 references that are in the map.
- First posted on arXiv 25 Feb 2026.

## Placed in

- World models / [World models for language agents](https://mlsanity.com/wm/world-models-for-language-agents/)

## Builds on (most cited first, 9 of 9)

- [World Models](https://mlsanity.com/paper/1803.10122/) (2018)
- [Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model](https://mlsanity.com/paper/1911.08265/) (2019)
- [Mathematical exploration and discovery at scale](https://mlsanity.com/paper/2511.02864/) (2025)
- [Generating Code World Models with Large Language Models Guided by Monte Carlo Tree Search](https://mlsanity.com/paper/2405.15383/) (2024)
- [Deep Reinforcement Learning Based Parameter Control in Differential Evolution](https://mlsanity.com/paper/1905.08006/) (2019)
- [Fast Genetic Algorithms](https://mlsanity.com/paper/1703.03334/) (2017)
- [Self-Adjusting Mutation Rates with Provably Optimal Success Rules](https://mlsanity.com/paper/1902.02588/) (2019)
- [On the Effectiveness of Simple Success-Based Parameter Selection Mechanisms for Two Classical Discrete Black-Box Optimization Benchmark Problems](https://mlsanity.com/paper/1803.01425/) (2018)
- [Architectures of Error: A Philosophical Inquiry into AI and Human Code Generation](https://mlsanity.com/paper/2505.19353/) (2025)

## Abstract (arXiv)

Can an LLM learn how an optimizer behaves -- and use that knowledge to control it? We extend Code World Models (CWMs), LLM-synthesized Python programs that predict environment dynamics, from deterministic games to stochastic combinatorial optimization. Given suboptimal trajectories of (1+1)-RLSk, the LLM synthesizes a simulator of the optimizer's dynamics; greedy planning over this simulator then selects the mutation strength k at each step. On lo and onemax, CWM-greedy performs within 6% of the theoretically optimal policy -- without ever seeing optimal-policy trajectories. On jumpk, where a deceptive valley causes all adaptive baselines to fail (0% success rate), CWM-greedy achieves 100% success rate -- without any collection policy using oracle knowledge of the gap parameter. On the NK-Landscape, where no closed-form model exists, CWM-greedy outperforms all baselines across fifteen independently generated instances (36.94 vs. 36.32; p<0.001) when the prompt includes empirical transition statistics. The CWM also outperforms DQN in sample efficiency (200 offline trajectories vs. 500 online episodes), success rate (100% vs. 58%), and generalization (k=3: 78% vs. 0%). Robustness experiments confirm stable synthesis across 5 independent runs.