It is found that neutral framing of behavioural features, reporting the feature value without prescriptive advice, consistently outperforms the prescriptive variants, and that prescriptive feedback steers the median feature value in the advised direction in only 37% of cases despite empirically grounded advice.
In evolutionary algorithms powered by language models, the LLM acts as a single operator that simultaneously updates structural components (like control flow) and continuous parameters. While LLMs can be good at the first, they are not efficient at the second, wasting tokens taking discrete jumps inside a trial and error loop. We resolve this by formalizing a hybrid nested search, in which an outer loop has the LLM propose a structural sketch, with numeric gaps, and an inner numerical optimizer tunes the sketch. Both the outer and inner solvers are pluggable: any text-based optimizer can be combined with a zero-order optimizer (CMA-ES), gradient-based routines, or MCMC samplers. We validate our framework across three scientific domains: (i) meta-optimizers on closed-form test functions, (ii) code-based policies for systems research and social dilemmas; and (iii) approximate Bayesian inference tasks. Across all three, the hybrid optimizer is superior to both vanilla LLM-driven search and pure numerical optimization baselines. Code at: https://github.com/vicgalle/hybrid-nested-search
Findings indicate that while LLM-based performance prediction is not yet a reliable substitute for benchmarking, it shows potential as a complementary pre-screening tool in iterative algorithm design, particularly in settings where code modifications follow predictable patterns.
Michal Pluháček, Paweł Kolendo, Krzysztof Tylka-Suleja et al.· Proceedings of the Genetic a...· 0 citations
In this entry to the GECCO 2026 Competition on LLM-Designed Metaheuristics, three large language models (GPT-5.4 Thinking, Gemini 3.1 Pro, Claude Opus 4.6) were independently prompted to design a metaheuristic for the GNBG benchmark (24 problems, hence the /24 scoring scale). Each reached a per-model ceiling (GPT 19.82, Claude 21.08, Gemini 23.29) and could not improve further on its own. Handing the best design (Gemini's) to the second-best model (Claude) for refinement produced, within four iterations, EO-BIPOP-CMA-CDR, which ranks first in the full 24-algorithm field at 23.30/24 and outperforms the source design head-to-head by three score points. A single cross-model handoff yielded more progress than five further self-refinement iterations by the source model. We argue that model diversity is a usable, low-cost resource for LLM-driven algorithm design.
Michal Pluháček, Paweł Kolendo, Krzysztof Tylka-Suleja et al.· Proceedings of the Genetic a...· 0 citations
Combinatorial problems appear in numerous industrial applications. A common approach is to formulate these problems as declarative constraint models that can subsequently be compiled to and solved by a range of back-end solvers. Recent work shows that Large Language Models (LLMs) can produce correct models from natural language, but even a correct model can be expensive to solve because performance remains sensitive to modelling choices. In this work, we investigate whether LLMs can automate performance-oriented model reformulation. Inspired by Automatic Heuristic Design (AHD), we use an evolutionary framework in which an LLM proposes candidate reformulations that are verified and benchmarked against the user-defined baseline model. We compare AHD-adapted search strategies that control which prior attempts, instructions, and measured feedback enter each prompt. Existing retention strategies prioritize recency or performance, but do not explicitly diversify the context. To cover this gap, we introduce Profile-Diverse Retention (PDR), which applies Maximal Marginal Relevance (MMR) to instance-level runtime vectors to retain behaviourally diverse attempts. We systematically evaluate the strategies on eight CSPLib problems using validation-based final model selection. The results show that: (i) iterative reformulation can produce substantial held-out speedups; (ii) strategies that keep the retained context diverse outperform those that retain only recent or the fastest attempts; and (iii) validation-based selection improves the held-out speedup of every strategy.
Kostis Michailidis, Dimos Tsouros, Dang Nguyen et al.· 0 citations
Analysis shows that many generated operators use semantics to guide selection, suggesting that LLMs can produce non-trivial search heuristics from the task description alone, and the relationship between public LLM leaderboard rankings and GP performance is examined.
Hengzhe Zhang, Qi Chen, Bing Xue et al.· 1 citation
Overall, statistic synthesis is much easier than map synthesis, some collections remain near-zero, long prompts cause a sharp accuracy cliff, and exact symbolic rule induction remains brittle.