LLMatic: Neural Architecture Search Via Large Language Models And Quality Diversity Optimization
Open accessThe DOI matches its Crossref record; 37 of the 42 references it cites were found in a registry or an index.
Abstract
Large language models (LLMs) have emerged as powerful tools capable of accomplishing a broad spectrum of tasks. Their abilities span numerous areas, and one area where they have made a significant impact is in the domain of code generation. Here, we propose using the coding abilities of LLMs to introduce meaningful variations to code defining neural networks. Meanwhile, Quality-Diversity (QD) algorithms are known to discover diverse and robust solutions. By merging the code-generating abilities of LLMs with the diversity and robustness of QD solutions, we introduce LLMatic, a Neural Architecture Search (NAS) algorithm. While LLMs struggle to conduct NAS directly through prompts, LLMatic uses a procedural approach, leveraging QD for prompts and network architecture to create diverse and high-performing networks. We test LLMatic on the CIFAR-10 and NAS-bench-201 benchmarks, demonstrating that it can produce competitive networks while evaluating just 2, 000 candidates, even without prior knowledge of the benchmark domain or exposure to any previous top-performing models for the benchmark. The open-sourced code is available at https://github.com/umair-nasir14/LLMatic.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Figures









Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×
“The intermediate residual block is the basic residual block with a stride of 2 (He et al., 2016), which serves to downsample the spatial size and double the channels of an input feature map.”
BackgroundVaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404 · cited 1×
“Transformers [6] facilitated the explosion of LLMs [7, 8].”
BackgroundImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 1×
“The specified cell within the search domain is depicted as a densely connected directed acyclic graph with four nodes and six ImageNet16-120 is a variant of ImageNet dataset [38] which is down-sampled to 16x16 image sizes and contains the first 120 classes.”
Multidimensional binary search trees used for associative searching
Bentley, J. L. (1975) · Communications of the ACM · doi:10.1145/361002.361007 · Cited by 7,520
Large-Scale Machine Learning with Stochastic Gradient Descent
Bottou, L. (2010) · Proceedings of COMPSTAT'2010 · doi:10.1007/978-3-7908-2604-3_16 · Cited by 5,795 · cited 1×
“The networks are optimized by stochastic gradient descent [21] with the learning rate set to 0.”
BackgroundMnasNet: Platform-Aware Neural Architecture Search for Mobile
Tan, M., Chen, B., Pang, R., Vasudevan, V., Sandler, M., Howard, A., & Le, Q. V. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00293 · Cited by 3,416 · cited 2×
“…a fixed number of nodes/layers and edges/operations) and the corresponding test accuracy of each architecture after training on a given dataset (e.g. CIFAR-10), allowing researchers to explore the tradeoffs of various NAS algorithms without needing to retrain each candidate network during search.”
BackgroundMethodEvolving Neural Networks through Augmenting Topologies
Stanley, K. O., & Miikkulainen, R. (2002) · Evolutionary Computation · doi:10.1162/106365602320169811 · Cited by 3,258 · cited 1×
“LLMatic begins its search with a very basic neural network, inspired by the work of Stanley [15] which suggests that neuroevolution tends to perform better when starting with a small network.”
DARTS: Differentiable architecture search
Liu, H., Simonyan, K., & Yang, Y. (2018) · arXiv · doi:10.48550/arxiv.1806.09055 · Cited by 3,248 · cited 1×
“We show empirically that LLMatic can find more than 20 competitive networks in CIFAR-10 and near-to-optimal networks in NAS-bench-201, using only 2000 evaluations.”
MethodTorrey, L., & Shavlik, J. (2010) · Handbook of Research on Machine Learning Applications and Trends · doi:10.4018/978-1-60566-766-9.ch011 · Cited by 1,503
Research works printed but not linked: 10
- Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural information processing systems 33 (2020), 1877-1901.
- Sergey Ioffe and Christian Szegedy. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In International conference on machine learning. pmlr, 448-456.
- Kirthevasan Kandasamy, Willie Neiswanger, Jeff Schneider, Barnabas Poczos, and Eric P Xing. 2018. Neural architecture search with bayesian optimisation and optimal transport. Advances in neural information processing systems 31 (2018).
- Vinod Nair and Geoffrey E Hinton. 2010. Rectified linear units improve re- stricted boltzmann machines. In Proceedings of the 27th international conference on machine learning (ICML-10). 807-814. · doi:10.5555/3104322.3104425
- Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. 2019. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems 32 (2019).
- Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. 2019. Language models are unsupervised multitask learners. OpenAI blog 1, 8 (2019), 9.
- Mingxing Tan and Quoc Le. 2019. Efficientnet: Rethinking model scaling for convolutional neural networks. In International conference on machine learning. PMLR, 6105-6114.
- Manoel Tenorio and Wei-Tsih Lee. 1988. Self organizing neural networks for the identification problem. Advances in Neural Information Processing Systems 1 (1988).
- Petar Veličković, Adrià Puigdomènech Badia, David Budden, Razvan Pascanu, Andrea Banino, Misha Dashevskiy, Raia Hadsell, and Charles Blundell. 2022. The CLRS algorithmic reasoning benchmark. In International Conference on Machine Learning. PMLR, 22084-22102.
- Chris Ying, Aaron Klein, Eric Christiansen, Esteban Real, Kevin Murphy, and Frank Hutter. 2019. Nas-bench-101: Towards reproducible neural architecture search. In International Conference on Machine Learning. PMLR, 7105-7114.
Cited by
Large Language Models for Software Engineering: A Systematic Literature Review
Hou, X., Zhao, Y., Liu, Y., Yang, Z., Wang, K., Li, L., Luo, X., Lo, D., Grundy, J., & Wang, H. (2024) · ACM Transactions on Software Engineering and Methodology · doi:10.1145/3695988 · Cited by 1,029
Mathematical discoveries from program search with large language models
Romera-Paredes, B., Barekatain, M., Novikov, A., Balog, M., Kumar, M. P., Dupont, E., Ruiz, F. J. R., Ellenberg, J. S., Wang, P., Fawzi, O., et al. (2024) · Nature · doi:10.1038/s41586-023-06924-6 · Cited by 544
“To surpass the ‘nominal’ capabilities of LLMs, recent studies 3 have combined them with evolutionary algorithms 14,15 , leading to important improvements on diverse synthetic problems 16 , searching for neural network architectures 17–19 and solving puzzles 20 .”
BackgroundWhen large language models meet personalization: perspectives of challenges and opportunities
Chen, J., Liu, Z., Huang, X., Wu, C., Liu, Q., Jiang, G., Pu, Y., Lei, Y., Chen, X., Wang, X., et al. (2024) · World Wide Web · doi:10.1007/s11280-024-01276-1 · Cited by 367
“LLMatic [177] and EvoPrompting [178] use codeLLMs as mutation and crossover operators for a genetic NAS algorithm.”
Evolutionary Computation in the Era of Large Language Model: Survey and Roadmap
Wu, X., Wu, S.-H., Wu, J., Feng, L., & Tan, K. C. (2024) · IEEE Transactions on Evolutionary Computation · doi:10.1109/tevc.2024.3506731 · Cited by 330
“2) With Code Generation Capability of General LLM: Nasir et al. [126] leveraged the code generation ability of large models to help NAS.”
Algorithm Evolution Using Large Language ModelPreprint
Liu, F., Tong, X., Yuan, M., & Zhang, Q. (2023) · arXiv (Cornell University) · doi:10.13140/rg.2.2.18894.43847 · Cited by 164
A survey on large language models for software engineering
Zhang, Q., Fang, C., Xie, Y., Zhang, Y., Yu, S., Sun, W., Yang, Y., & Chen, Z. (2026) · Science China Information Sciences · doi:10.1007/s11432-025-4670-0 · Cited by 124
Large Language Models for Software Engineering: A Systematic Literature Review
Hou, X., Zhao, Y., Liu, Y., Yang, Z., Wang, K., Li, L., Luo, X., Lo, D., Grundy, J., & Wang, H. (2023) · arXiv · doi:10.48550/arxiv.2308.10620 · Cited by 114
A Systematic Survey on Large Language Models for Algorithm Design
Liu, F., Yao, Y., Guo, P., Yang, Z., Lin, X., Zhao, Z., Tong, X., Mao, K., Lu, Z., Wang, Z., et al. (2026) · ACM Computing Surveys · doi:10.1145/3787585 · Cited by 169
“Nasir et al. [110] introduce a quality-diversity algorithm tailored to NAS, producing architectures for CIFAR-10 and NAS-bench-201 benchmarks.”
When large language model meets optimization
Huang, S., Yang, K., Qi, S., & Wang, R. (2024) · Swarm and Evolutionary Computation · doi:10.1016/j.swevo.2024.101663 · Cited by 113
Evolving code with a large language model
Hemberg, E., Moskal, S., & O’Reilly, U.-M. (2024) · Genetic Programming and Evolvable Machines · doi:10.1007/s10710-024-09494-2 · Cited by 126
“In other NAS work Nasir et al. [40] use an LLM and a Quality-Diversity algorithm to obtain variations of Python code defining a neural architecture.”
Language Model Crossover: Variation through Few-Shot Prompting
Meyerson, E., Nelson, M. J., Bradley, H., Gaier, A., Moradi, A., Hoover, A. K., & Lehman, J. (2024) · ACM Transactions on Evolutionary Learning and Optimization · doi:10.1145/3694791 · Cited by 75
Language model crossover: Variation through few-shot prompting
Meyerson, E., Nelson, M. J., Bradley, H., Gaier, A., Moradi, A., Hoover, A. K., & Lehman, J. (2023) · arXiv · doi:10.48550/arxiv.2302.12170 · Cited by 75
Nie, T., Sun, J., & Ma, W. (2025) · Artificial Intelligence for Transportation · doi:10.1016/j.ait.2025.100003 · Cited by 116
“…and planning, world models allow autonomous systems to respond quickly and intelligently to complex and often unpredictable situations in a closed-loop manner (Gao et al., 2023b; Hu et al., 2023; Wang et al., 2024g,i; Gao et al., 2024b; Yang et al., 2024b; Zheng et al., 2024a; Fu et al.,…”
How Far Are We From AGI: Are LLMs All We Need?
Feng, T., Jin, C., Liu, J., Zhu, K., Tu, H., Cheng, Z., Lin, G., & You, J. (2024) · Transactions on machine learning research · doi:10.48550/arxiv.2405.10313 · Cited by 94
AutoML in the Age of Large Language Models: Current Challenges, Future Opportunities and Risks
Tornede, A., Deng, D., Eimer, T., Giovanelli, J., Mohan, A., Ruhkopf, T., Segel, S., Theodorakopoulos, D., Tornede, T., Wachsmuth, H., et al. (2024) · Transactions on machine learning research · doi:10.48550/arxiv.2306.08107 · Cited by 82
“Although there is an ongoing discussion in the community on what constitutes a good explanation in general, textual explanations seem to be highly trusted by users (Gilpin et al., 2018).”
BackgroundEvaluation of Large Language Models as Solution Generators in Complex Optimization
Huang, B., Wu, X., Zhou, Y. F., Wu, J., Feng, L., Cheng, R., & Tan, K. C. (2025) · IEEE Computational Intelligence Magazine · doi:10.1109/mci.2025.3580520 · Cited by 63
“Liu et al. [16] integrated the LLM as an operator within a decomposition-based multi-objective optimization framework, while Bradley et al. [17] integrated LLM Quality-Diversity (QD) search [36] and proposed QD through AI Feedback (QDAIF) algorithm.”
When Large Language Models Meet Evolutionary Algorithms: Potential Enhancements and Challenges
Wang, C., Zhao, J., Jiao, L., Li, L., Liu, F., & Yang, S. (2025) · Research · doi:10.34133/research.0646 · Cited by 76
“For example, in automated learning, training logs act as fitness for finding efficient architectures and hyperparameters [ 106 – 110 ].”
EDT: Improving Large Language Models' Generation by Entropy-based Dynamic Temperature SamplingPreprint
Zhang, S., Bao, Y., & Huang, S. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2403.14541 · Cited by 72
“While temperature sampling strategy has also been widely used to strike a balance between generation quality and diversity (Nasir et al., 2023), we have observed significant limitations in fixed temperature settings.”
BackgroundAdvances in neural architecture search
Wang, X. Y., & Zhu, W. (2024) · National Science Review · doi:10.1093/nsr/nwae282 · Cited by 70
“Besides, it is worth studying using the coding abilities of LLMs to introduce meaningful variations to code-defining neural network architecture [133 ].”
MethodHavrilla, A., Dai, A. M., O'Mahony, L., Oostermeijer, K., Zisler, V., Albalak, A., Milo, F., Raparthy, S. C., Gandhi, K., Abbasi, B., et al. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2412.02980 · Cited by 60
Fundamental Capabilities and Applications of Large Language Models: A Survey
Li, J., Gao, Y., Yang, Y., Bai, Y., Zhou, X., Li, Y., Sun, H., Liu, Y., Si, X., Ye, Y., et al. (2025) · ACM Computing Surveys · doi:10.1145/3735632 · Cited by 81
“For instance, WebGPT interacts with a text-based web browsing environment, enabling end-to-end optimization search and aggregation through imitation and reinforcement learning.”
Word2World: Generating Stories and Worlds through Large Language ModelsPreprint
Nasir, M. U., James, S. D., & Togelius, J. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2405.06686 · Cited by 48
“…and semantically rich, have not only redefined the boundaries of natu-ral language processing (Liu et al. 2023) but have also shown unprecedented versatility in tasks far beyond text generation, such as generating neural networks (Nasir et al. 2023b) and folding proteins (Lin et al. 2022).”
BackgroundSelf-citationPractical PCG Through Large Language Models
Nasir, M. U., & Togelius, J. (2023) · IEEE Conference on Games (CoG) · doi:10.1109/cog57401.2023.10333197 · Cited by 50
“It’s 2023, and the new new thing that can be applied to everything under the sun is Large Language Models (LLMs), such as image generation [8] and neural architecture search [9].”
BackgroundSelf-citationNeural Architecture Discovery via Autonomous EvolutionPreprint
Xu, W., Liu, Y., Nan, Y., Ye, L., Hu, X., Qin, Z., & Liu, P. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.18074 · Cited by 31
“Leveraging modern LLMs (Brown et al., 2020; OpenAI, 2023), ASI-A RCH autonomously hypothesizes novel architectural concepts, implements them as code, and validates performance through rigorous experimentation (Nasir et al., 2023; Cheng et al., 2025; Chen et al., 2025).”
Design Principle Transfer in Neural Architecture Search via Large Language Models
Zhou, X., Wu, X., Feng, L., Lu, Z., & Tan, K. C. (2025) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v39i21.34463 · Cited by 59
LLMs and IoT: A Comprehensive Survey on Large Language Models and the Internet of Things
Sarhaddi, F., Nguyen, N. T., Zuniga, A., Khan, M., Hui, P., Tarkoma, S., Flores, H., & Nurmi, P. (2026) · IEEE Open Journal of the Communications Society · doi:10.1109/ojcoms.2026.3680064 · Cited by 33
“…option for optimizing model efficiency is to use neural architecture search (NAS) techniques to identify a smaller model architecture that closely resembles the original while meeting the resource constraints of the target platform, with promising results reported in applying NAS to LLMs [96].”
Hallucination Reduction and Optimization for Large Language Model-Based Autonomous Driving
Wang, J. (2024) · Symmetry · doi:10.3390/sym16091196 · Cited by 25
Position: Leverage Foundational Models for Black-Box OptimizationPreprint
Song, X., Tian, Y., Lange, R. T., Lee, C., Tang, Y., & Chen, Y. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2405.03547 · Cited by 19
“(Chen et al., 2023; Nasir et al., 2023) use an LLM to adaptively mutate and cross-over code for evolutionary Neural Architecture Search.”
MethodOStr-DARTS: Differentiable Neural Architecture Search Based on Operation Strength
Yang, L., Zheng, Z., Han, Y., Song, S., Huang, G. M., & Li, F. (2024) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2024.3455760 · Cited by 21
“Recent studies have also achieved NAS using multiagent [35] or large language model (LLM) [36] for searching.”
BackgroundFunBO: Discovering Acquisition Functions for Bayesian Optimization with FunSearchPreprint
Aglietti, V., Ktena, S. I., Schrouff, J., Sgouritsa, E., Ruiz, F. J. R., Malek, A., Bellot, A., & Chiappa, S. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2406.04824 · Cited by 23
“We do not scale the output values nor normalize the input domains to facilitate learning, but rather use the objective functions as available in standard BO packages out-of-the-box.”
Evolution of Heuristics: Towards Efficient Automatic Algorithm Design Using Large Language Model
Liu, F., Tong, X., Yuan, M., Lin, X., Luo, F., Wang, Z., Lu, Z., & Zhang, Q. (2024) · arXiv · doi:10.48550/arxiv.2401.02051 · Cited by 14
“…et al., 2023; Brownlee et al., 2023; Liu et al., 2023c), multi-objective evolutionary algorithms Liu et al. (2023a), neural architecture design (Nasir et al., 2023; Jawahar et al., 2023; Chen et al., 2023), Monte Carlo Tree Search (Zhao et al., 2023b), graph-based combinatorial optimization…”
Pretrained Optimization Model for Zero-Shot Black Box Optimization
Li, X., Wu, K., Liu, J., Zhang, X., Wang, H., & Liu, J. (2024) · neural information processing systems · doi:10.52202/079017-0456 · Cited by 13
“In line with POMs, various optimization approaches leveraging Large Language Models (LLMs) have emerged to address diverse problem domains, including NP-hard problems [28, 29], algorithm evolution [30–33], reward design [34], and Neural Architecture Search (NAS) [35, 36].”
ACES: Generating Diverse Programming Puzzles with with Autotelic Generative ModelsPreprint
Pourcel, J., Colas, C., Molinaro, G., Oudeyer, P., & Teodorescu, L. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2310.10692 · Cited by 18
“Their capacity to generate appropriate variations of existing text has been used to implement mutation [ELM: Lehman et al., 2022] and crossover [Meyerson et al., 2023] operators within QD, with applications to neural architecture search [Chen et al., 2023, Nasir et al., 2023].”
B2Opt: Learning to Optimize Black-box Optimization with Little Budget
Li, X., Wu, K., Zhang, X., & Wang, H. (2025) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v39i17.34036 · Cited by 11
On neural architecture search and hyperparameter optimization: A max-flow based approach
Xue, C., Li, J., Wang, X., Zhan, Y., Yan, J., & Li, C.-G. (2025) · Neural Networks · doi:10.1016/j.neunet.2025.107507 · Cited by 10
$T^2$ of Thoughts: Temperature Tree Elicits Reasoning in Large Language ModelsPreprint
Cai, C., Zhao, X., Du, Y., Liu, H., & Li, L. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2405.14075 · Cited by 10
Grammar-Based Game Description Generation Using Large Language Models
Tanaka, T., & Simo‐Serra, E. (2024) · IEEE Transactions on Games · doi:10.1109/tg.2024.3520214 · Cited by 12
“Quality-diversity algorithms utilizing LLMs have been proposed for controllable program synthesis, such as neural architecture search [57].”
Duvvuri, R., Vysyaraju, C., Goyal, A., Ignatov, D., & Timofte, R. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2512.24120 · Cited by 10
“Direct quantitative comparison with EvoPrompting [4] and LLMatic [34] is challenging due to different evaluation protocols: EvoPrompting eval-6 While EvoPrompting demonstrates that evolutionary prompt optimization can improve generation quality, our finding that n = 3 provides stable performance…”
InfluentialFrom Text to Life: On the Reciprocal Relationship between Artificial Life and Large Language Models
Nisioti, E., Glanois, C., Najarro, E., Dai, A. M., Meyerson, E., Pedersen, J. W., Teodorescu, L., Hayes, C. F., Sudhakaran, S., & Risi, S. (2024) · Conference on Artificial Life (ALIFE) · doi:10.1162/isal_a_00759 · Cited by 18
Nasir, M. U., James, S. D., & Togelius, J. (2024) · neural information processing systems · doi:10.52202/079017-1000 · Cited by 17
Self-citationCan LLMs Revolutionize the Design of Explainable and Efficient TinyML Models?
Zeinaty, C. E., Hamidouche, W., Herrou, G., Ménard, D., & Debbah, M. A. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11229057 · Cited by 10
“Unlike traditional NAS methods, which often rely on computationally intensive strategies such as reinforcement learning or evolutionary algorithms [25], [26], LLMs enable a more rapid and flexible exploration of the design space through their inherent reasoning and generative capabilities [11], [12].”
Video game development 3.0: AI-driven collaborative co-creation
Ratican, J., & Hutson, J. (2024) · Metaverse · doi:10.54517/m2904 · Cited by 10
LLM-Guided Evolution: An Autonomous Model Optimization for Object Detection
Yu, Y., & Zutty, J. (2025) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3712255.3734340 · Cited by 10
“Recently, this type of approach has been demonstrated several times on common benchmark problems such as image classification with the CIFAR-10 dataset [14, 15], MNIST [2], and the traveling salesman problem [13].”
A Survey on Evolutionary Multimodal Optimization-Driven Machine Learning
Wang, P., Xue, B., Liang, J., & Zhang, M. (2025) · IEEE Transactions on Evolutionary Computation · doi:10.1109/tevc.2025.3627487 · Cited by 6
“Nasir et al. [207] employed LLMs to generate novel architectural variants integrated with an QD optimization algorithm (named MAP-Elites) to discover diverse and robust solutions.”
Survey of Genetic Programming and Large Language Models
Hemberg, E., Jorgensen, S., & O’Reilly, U.-M. (2025) · Genetic and evolutionary computation · doi:10.1007/978-981-96-0077-9_4 · Cited by 7
Language Model Evolutionary Algorithms for Recommender Systems: Benchmarks and Algorithm Comparisons
Liu, J., Sun, Z., Feng, S., Chen, C., & Ong, Y.-S. (2025) · IEEE Transactions on Evolutionary Computation · doi:10.1109/tevc.2025.3609058 · Cited by 8
“Restrictions apply. ture search [15], and code generation [16].”
Solve it with EASEPreprint
Viktorin, A., Kadavý, T., Kováč, J., Žáček, P., Malaník, D., Martinek, T., Pluháček, M., & Šenkeřík, R. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7465221 · Cited by 5
Delta-Based Neural Architecture Search: LLM Fine-Tuning via Code DiffsPreprint
Adhikari, S. P., Timofte, R., & Ignatov, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.04903 · Cited by 5
“The recent advent of code-capable LLMs [6, 14, 18, 41, 45] has opened an alternative avenue: using language models as architecture generators that produce executable PyTorch code directly [3, 24, 35].”
LLMENAS: Evolutionary Neural Architecture Search via Large Language Model Guidance
Lai, Y., Cai, Z., Chen, L., Ling, T., & Liu, H. (2026) · IEEE Transactions on Evolutionary Computation · doi:10.1109/tevc.2026.3670336 · Cited by 4
“Nasir et al. [15] proposed a method called LLMatic, which combines the generative capabilities of LLMs with the diversity and robustness of quality diversity (QD) algorithms to search for optimal architectures.”
Closed-Loop LLM Discovery of Non-standard Channel Priors in Vision Models
Uzun, T. A., Ignatov, D., & Timofte, R. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-31920-3_42 · Cited by 8
LLM-NAS: LLM-driven Hardware-Aware Neural Architecture Search
Zhu, H., Zhang, G. L., & Huang, S. (2026) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3795101.3805321 · Cited by 4
“Recently, Large Language Models (LLMs) offer a promising training-free alternative for discovering neural architectures Achiam et al. (2023).”
AgentSquare: Automatic LLM Agent Search in Modular Design Space
Shang, Y., Li, Y., Zhao, K., Ma, L., Liu, J., Xu, F., & Li, Y. (2024) · arXiv · doi:10.48550/arxiv.2410.06153 · Cited by 3
“2.1 B ACKGROUND Using LLMs for automatic optimization has been a widely explored topic, such as applications in code generation (Lehman et al., 2023; Romera-Paredes et al., 2024) and neural architecture search (Nasir et al., 2024; Chen et al., 2024a).”
InfluentialNADER: Neural Architecture Design via Multi-Agent Collaboration
Yang, Z., Zeng, W., Jin, S., Chen, Q., Luo, P., & Liu, W. (2025) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52734.2025.00420 · Cited by 12
Llm-ga: A gradient-based multi-label adversarial attack by large language models
Liu, Y., Hu, Y., Chen, Z., Wang, S., & Luo, W. (2025) · Complex & Intelligent Systems · doi:10.1007/s40747-025-02184-1 · Cited by 3
Diffusion evolutionary algorithm for neural architecture search
Xue, Y., Cai, P., Neri, F., Zhang, Y. Q., & Gong, D. (2026) · Swarm and Evolutionary Computation · doi:10.1016/j.swevo.2026.102421 · Cited by 3
LLM-Guided Evolution: An Autonomous Model Optimization for Object DetectionPreprint
Yu, Y., & Zutty, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.02280 · Cited by 3
“Recently, this type of approach has been demonstrated several times on common benchmark problems such as image classification with the CIFAR-10 dataset [14, 15], MNIST [2], and the traveling salesman problem [13].”
Agentic Neural Architecture SearchPreprint
Jeong, S., Kim, M., & Kim, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.07984 · Cited by 4
“…literature, they can generate or mutate network code in an open-ended space, serving as mutation operators (Chen et al., 2023; Morris et al., 2024; Nasir et al., 2024; Zhu et al., 2025), optimizers (Zheng et al., 2023; Yu et al., 2023), and hyperparameter tuners (Zhang et al., 2023; Liu et al.,…”
Foundation model-based generative AI for smart manufacturing: A paradigm shift
Zheng, H., Geng, Z., & Xu, X. W. (2026) · ENGINEERING Management · doi:10.1007/s42524-026-5359-0 · Cited by 6
“Nasir et al. (2024) • Explainability and robustness: Beyond performance, GenAI addresses the black-box nature of AI.”
EvoDiff-NAS: Evolutionary neural architecture search with diffusion-guided mutation
Tang, Z., Xie, Z., Wen, X., Tang, B., & Song, W. (2026) · Swarm and Evolutionary Computation · doi:10.1016/j.swevo.2026.102442 · Cited by 2
Using Large Language Models for Evolutionary Search
O’Reilly, U.-M., & Hemberg, E. (2024) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3638530.3648432 · Cited by 2
Katasani, C. (2025) · International Journal of Scientific Research in Computer Science Engineering and Information Technology · doi:10.32628/cseit25111263 · Cited by 2
Web Personalization with Large Language Models: Challenges and Future Trends
Bansal, N., Bala, M., & Sharma, K. (2025) · Algorithms for intelligent systems · doi:10.1007/978-981-97-8669-5_21 · Cited by 3
Automated UAV Object Detector Design Using Large Language Model-Guided Architecture Search
Kong, F., Shan, X., Hu, Y., & Li, J. (2025) · Drones · doi:10.3390/drones9110803 · Cited by 2
“Recent works start to employ LLMs to (i) propose architecture blueprints from natural-language constraints, (ii) synthesize runnable code, and (iii) iterate with scoring feed-back [18–22].”
Augmentative Topology Agents For Open-Ended Learning
Nasir, M. U., Beukman, M., James, S. D., & Cleghorn, C. W. (2023) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3583133.3590576 · Cited by 4
“QD algorithms may also be worthwhile to explore in the context of open-ended learning as they have the ability to produce a population of high-performing and diverse individuals [68, 69].”
BackgroundSelf-citationAn AI4AI Framework for Visual Token PruningPreprint
Liu, Z., Huang, W., Song, W., Liu, Y., Yang, Z., & Fu, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.07193 · Cited by 2
“In neural architecture search, EvoPrompt-ing (Chen, Dohan, and So 2023) and LLMatic (Nasir et al. 2023) employ LLMs as architecture generators or mutation operators.”
EA-HAS-Bench and Language-Enhanced Shrinkage Search for Energy-Aware NAS
Zhao, C., Dou, S., Zhao, J., Jiang, X., Gao, J., Zhang, Y., Li, B., & Li, D. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3598206 · Cited by 1
“The first approach leverages LLMs for generating or recommending architectural frameworks, with significant contributions from LLMatic [42] and EvoPrompting [43], which explore LLMs’ potential to autonomously create new architectures.”
A General Genetic Algorithm Using Natural Language Evolutionary Operators
Stenzel, G., Gerner, S., Kölle, M., Zorn, M., & Gabor, T. (2025) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3712255.3726695 · Cited by 1
“On a more astract level, architecture search, i.e., finding the right combination of components for the evolutionary algorithm [14] or for a neural network [4, 12] (which the EA is then meant to tune), has yielded various successes.”
Gao, F., Zhao, C., Gong, D., Yan, Z., Zhao, H., & Zheng, Q. (2025) · Circuits Systems and Signal Processing · doi:10.1007/s00034-025-03415-z · Cited by 1
“LLMs are deep neural network architectures [ 30] with billions of parameters, trained on massive datasets to achieve strong performance in natural language processing, knowledge reasoning, and autonomous generation.”
Rojas, L., Vega, E., Jorquera, L., & García, J. (2026) · Mathematics · doi:10.3390/math14071091 · Cited by 2
LLMs for End-to-End Machine Learning: A Comprehensive Review
Takallou, M. A., Gautam, A., & Shirolkar, P. (2026) · IEEE Access · doi:10.1109/access.2026.3678429 · Cited by 2
PhaseNAS: Language-Model Driven Architecture Search with Dynamic Phase Adaptation
Kong, F., Shan, X., Hu, Y., & Li, J. (2026) · Lecture notes in electrical engineering · doi:10.1007/978-981-95-7656-2_47 · Cited by 1
Mortar: Evolving Mechanics for Automatic Game Design
Nasir, M. U., Li, Y., James, S. D., & Togelius, J. (2026) · Genetic and Evolutionary Computation Conference (GECCO) · doi:10.1145/3795095.3805100 · Cited by 3
“This approach has been applied to robot morphology evolution [14], and neural architecture search [19].”
Self-citationLyria: A Genetic Algorithm-Driven Neuro-Symbolic Reasoning Framework for LLMsPreprint
Tang, W., Nuamah, K., & Belle, V. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.04034 · Cited by 2
What an Autonomous Agent Discovers About Molecular Transformer Design: Does It Transfer?Preprint
Wijaya, E. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.28015 · Cited by 1
“ELM [Lehman et al., 2022] and LLMatic [Nasir et al., 2023] explore evolution-through-LLMs and quality-diversity optimization respectively.”
Project Auto-World: Towards Automated Benchmarking of Neural Relational ReasonersPreprint
Das, A., Boisson, J., Khalid, I., Garai, S., & Schockaert, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.24965 · Cited by 1
“Within a wider setting, LLM-based evolution has been applied to prompt engineering [Guo et al., 2024, Agrawal et al., 2026], neural architecture search [Nasir et al., 2024] and reward shaping [Hazra et al., 2025, Sapora et al., 2024].”
Knowledgeable Language Models as Black-Box Optimizers for Personalized Medicine
Yao, M. S., Bastani, O., Andersson, A., Biancalani, T., BenTaieb, A., & Iriondo, C. (2025) · PubMed · doi:10.48550/arxiv.2509.20975
Cascade Adversarial Attack Search
Wang, Z., Shen, D., Sun, X., Yang, S., Yu, X., & Tian, Y. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11228300
Language Models Speed Up Local Search for Finding Programmatic Policies
Sadmine, Q. A., Baier, H., & Lelis, L. H. S. (2024) · Transactions on machine learning research · doi:10.5281/zenodo.14187699 · Cited by 3
On the use of LLMs to generate a dataset of Neural NetworksPreprint
Daoudi, N., & Cabot, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.04388
“LLMs have been used for NAS in several works [13, 4, 21, 30].”
Sharon Roji Priya, C., Perumalsamy, D., & Thinakaran, R. (2026) · Future Internet · doi:10.3390/fi18030132
“In modern language models, hyperparameter spaces remain high-dimensional [32].”
Learning to Grasp in Robotics with Quality-Diversity Generated Data
Huber, J. B. (2024) · theses.fr (ABES) · doi:10.70675/b98f392dz4df9z4646z8b34z59f447564543
Neural architecture search from a natural language processing perspective: a survey
Yan, Y., Mo, L., & Tan, Y. (2026) · Artificial Intelligence Review · doi:10.1007/s10462-026-11550-5
“LLMatic (Nasir et al. 2024) embeds LLMs as evolutionary operators to generate mutations informed by successful design patterns, while GPT-NAS (Yu et al. 2025) reformulates NAS as sequence modeling, pre-training GPT models to autoregressively generate novel designs.”
HERCULES: Hardware-Efficient, Robust, Continual Learning Neural Architecture SearchPreprint
Gambella, M., Pittorino, F., & Roveri, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.04103
“Another opportunity lies in foundation models for NAS, where pretrained architecture generators learn transferable structural priors to rapidly adapt to new tasks, hardware constraints, or robustness requirements [159], [160].”
Jeya, J. S., Arora, L., Sinha, A., Jain, T., Tiwari, D., & Sharma, O. (2026) · International Conference on Research Methodologies in Knowledge Management, Artificial Intelligence and Telecommunication Engineering (RMKMATE) · doi:10.1109/rmkmate69073.2026.11518979
“LLMatic demonstrated that LLMs can generate architectural variants and be integrated into quality-diversity optimization [9].”
AbuSadeh, M., Wei, L., & Zhang, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.30109
“In model and architecture discovery, QD provides a principled way to retain multiple high-quality architectural motifs while exploring trade-offs instead of converging prematurely to one design family [25].”
Curating Same-Family Neural Networks for LLM-Guided Model Improvement: A Controlled Case StudyPreprint
Baghel, K. K. S., Timofte, R., & Ignatov, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.05704
“OPRO, EvoPrompt, and LLMatic place LLMs inside optimization loops [10, 20, 26].”
LLM-Guided Run-Time Parameter Optimization for Energy-Efficient Model Inference
Crumpacker, M. K. (2026) · VTechWorks (Virginia Tech)
LEED: LLM-Empowered Event Denoising with Hardware Awareness
Shen, Z., Zhong, Y., Guo, S., Yang, Z., Wang, L., Luo, L., & Yuan, J. (2026) · Communications in computer and information science · doi:10.1007/978-981-92-3541-4_35
A dataset of neural network architectures generated via large language models
Daoudi, N., & Cabot, J. (2026) · Scientific Data · doi:10.1038/s41597-026-07893-x
Scaling Closed-Loop Feature Channel Configuration with LLMsPreprint
Uzun, T. A., Timofte, R., & Ignatov, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.20516
“EvoPrompting and LLMatic use language-model code generation in evolutionary or quality-diversity search loops [4, 14].”
RE-MMNAS: LLM Reasoning-Guided Evolutionary Architecture Search for Multi-Modal Feature Fusion
Xiao, H., Yan, X., Xie, Y. H., & Huang, H. P. (2026) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3795101.3814657
AutoEJS: Efficient and Effective Joint Hyperparameter and Architecture Search
Yan, Y., Wang, B., Yang, B., Chen, X., Liu, J., & Wang, Q. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-36217-9_22
End-to-End Model Generation with Large Language Models for Adaptive IoT Application Deployment
Wen, Z., Feng, J., Yao, N., Wu, D., Wang, C., Wu, M., Qin, J., & He, S. (2026) · ACM/IEEE International Conference on Software Engineering: Companion Proceedings (ICSE-Companion) · doi:10.1145/3744916.3787771
Prompt-Guided Resource-Aware Neural Architecture Search for TinyML Object Detection Models
Sakib, N., Zennaro, M., & Nower, N. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-36868-3_18
Peng, B., Zhang, L., Zhou, Y., Wang, M., Wu, Y., Li, S., Fan, X., & Zhao, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.05950
“LLM-guided search enables flexible code edits (Romera-Paredes et al., 2024; Liu et al., 2024; Novikov et al., 2025; Lange et al., 2025), while quality-diversity methods maintain diverse candidates (Mouret & Clune, 2015; Nasir et al., 2024).”
EQD-Guard: Robust Post-Hoc Ensemble Selection with Rigorous Performance Guard
Zhou, J., Yu, R., & Wang, Y. (2026) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn69854.2026.11728449
A Task-Aware Rank-Based Trained LLM Surrogate for Cross-Task Neural Architecture Search
Wu, K.-C., Huang, H.-Q., Wang, H.-Y., & Tsai, C. (2026) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn69854.2026.11729119
“Early LLM-based studies, such as GENIUS and LLMatic, primarily employed LLMs as assistants within existing search pipelines [19]–[21].”
QuaNAS: Uncertainty-Aware Reuse of Earlier Searches for Query-Efficient Neural Architecture SearchPreprint
Satria, B. S. P., Sunaryono, D., Subakti, M. M. I., Sabilla, I. A., & Suciningtyas, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.31251
“Recent generators that propose architectures with large language models or diffusion models (Chen et al., 2023; Nasir et al., 2024; An et al., 2024) inherit the same evaluation questions.”
Londt, S., Gao, X., & McCallum, S. (2026) · IEEE Congress on Evolutionary Computation (CEC) · doi:10.1109/cec69861.2026.11727383
“Several LLM-augmented evolutionary methods rely on few-shot prompting or explicit training sets to guide model behaviour [8], [9], [10].”
Metadata
- Authors
- Nasir, M. U., Earle, S., Togelius, J., James, S., & Cleghorn, C.
- Affiliations
- University of the Witwatersrand (ZA)
- Amazon (United States) (US)
- Venue
- Proceedings of the Genetic and Evolutionary Computation Conference
- Publisher
- ACM
- Published
- 2024-07-14
- Pages
- 1110-1118
- OpenAlex
- W4400413686
- Cited by
- 52
- References
- 18
- Citations by year
- 2026: 21 · 2025: 18 · 2024: 11
- FWCI
- 14.76
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- Genetic and Evolutionary Computation Conference (GECCO) (publishedVersion)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- Center for High Performance Computing
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Evolutionary Algorithms and Applications
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Data Classification
- Keywords
- neural architecture search, large language models, code generation, procedural content generation, robustness, NAS-Bench-201, CIFAR-10
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.