mlsanityData to 7 Oct 2026

Code World Models for Parameter Control in Evolutionary Algorithms

arXiv 2602.22260 ↗First posted 25 Feb 2026cs.LG cs.NEcited by 0

On the map: World models → World models for language agents

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.

Key facts. No citing papers in the map yet (data through 7 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.

Builds on

9 of 24 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Generating Code World Models with Large Language Models Guided by Monte Carlo Tree Search24 May

Other references · 8

  1. 2018World Modelscited by 1,444
  2. 2019Mastering Atari, Go, Chess and Shogi by Planning with a Learned Modelcited by 1,327
  3. 2025Mathematical exploration and discovery at scalecited by 66
  4. 2019Deep Reinforcement Learning Based Parameter Control in Differential Evolutioncited by 37
  5. 2017Fast Genetic Algorithmscited by 29
  6. 2019Self-Adjusting Mutation Rates with Provably Optimal Success Rulescited by 13
  7. 2018On the Effectiveness of Simple Success-Based Parameter Selection Mechanisms for Two Classical Discrete Black-Box Optimization Benchmark Problemscited by 1
  8. 2025Architectures of Error: A Philosophical Inquiry into AI and Human Code Generationcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025CWM: An Open-Weights LLM for Research on Code Generation with World Models
  2. 2026When a Verified World Model Still Loses: Play-Adequacy vs Prediction-Accuracy in LLM-Synthesized Code World Models
  3. 2024WorldCoder, a Model-Based LLM Agent: Building World Models by Writing Code and Interacting with the Environment
  4. 2026Distilling Game Code World Model Generation into Lightweight Large Language Models
  5. 2026Code to Control: Synthesizing Parameterized Reactive Controllers
  6. 2026PatchWorld: Gradient-Free Optimization of Executable World Models for Agent Environments
  7. 2026Belief-Calibrated Optimization: An Explicit World Model for Agentic Optimization
  8. 2025Code World Models for General Game Playing