Mutation Without Variation: Convergence Dynamics in LLM-Driven Program Evolution
Quick Answer
This paper shows that LLM-driven program mutations show significant convergence, with 87% of mutation chains revisiting structural forms.
Quick Take
This structural bias limits open-ended exploration, highlighting a tension in capabilities. The study reveals that variations are mostly confined to terminal substitutions within recurring templates.
Key Points
- 87% of mutation chains revisit structural forms, indicating high convergence.
- Most variation occurs in terminal substitutions within recurring templates.
- Convergence rate varies with prompt wording and model choice.
- Classical GP subtree mutation does not show similar convergence patterns.
- Findings suggest a bias toward structural homogeneity in LLM mutations.
Paper Resources
📖 Reader Mode
~2 min readAbstract:When an LLM repeatedly mutates a program, does it explore new forms or circle back to the same ones? We study this question by analyzing LLM-driven mutation chains in the absence of selection pressure within a domain-specific language, varying prompt design, model family, and stochastic replication. We find that LLM-based mutation consistently converges toward restricted attractor regions in program space. Convergence is especially severe at the structural level: in 87% of chains, over 93% of mutations revisit a previously seen structural form, with most variation confined to terminal substitutions within recurring templates. Cycle analysis reveals short cycles and self-loops dominating the transition structure. The rate of convergence varies with prompt wording and model choice, but the phenomenon is robust across conditions. A classical GP subtree mutation operator does not exhibit comparable convergence, suggesting that the effect is intrinsic to the LLM mutation pipeline. These findings reveal a tension at the heart of LLM-driven program evolution: the same capabilities that enable semantics-aware program transformation also carry a systematic bias toward structural homogeneity that must be accounted for if such systems are to sustain open-ended exploration. Source code is available at this https URL.
| Comments: | Accepted to the Genetic and Evolutionary Computation Conference (GECCO '26) Workshop on Large Language Models for and with Evolutionary Computation |
| Subjects: | Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE) |
| Cite as: | arXiv:2606.05408 [cs.AI] |
| (or arXiv:2606.05408v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2606.05408 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Can Gurkan [view email]
[v1]
Wed, 3 Jun 2026 20:22:29 UTC (2,269 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.