# From Memorization to Creativity: LLM as a Designer of Novel Neural Architectures

arXiv: https://arxiv.org/abs/2601.02997 · HTML: https://mlsanity.com/paper/2601.02997/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Builds on 26 of its 42 references that are in the map.
- First posted on arXiv 6 Jan 2026.

## Builds on (most cited first, 26 of 26)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Neural Architecture Search with Reinforcement Learning](https://mlsanity.com/paper/1611.01578/) (2016)
- [DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence](https://mlsanity.com/paper/2401.14196/) (2024)
- [A Survey on Large Language Models for Code Generation](https://mlsanity.com/paper/2406.00515/) (2024)
- [Neural Architecture Search without Training](https://mlsanity.com/paper/2006.04647/) (2020)
- [Zero-Cost Proxies for Lightweight NAS](https://mlsanity.com/paper/2101.08134/) (2021)
- [Understanding and Robustifying Differentiable Architecture Search](https://mlsanity.com/paper/1909.09656/) (2019)
- [Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective](https://mlsanity.com/paper/2102.11535/) (2021)
- [Neural Architecture Search: Insights from 1000 Papers](https://mlsanity.com/paper/2301.08727/) (2023)
- [A Note on Shumailov et al. (2024): `AI Models Collapse When Trained on Recursively Generated Data'](https://mlsanity.com/paper/2410.12954/) (2024)
- [LLMatic: Neural Architecture Search via Large Language Models and Quality Diversity Optimization](https://mlsanity.com/paper/2306.01102/) (2023)
- [Self-Training: A Survey](https://mlsanity.com/paper/2202.12040/) (2022)
- [Zen-NAS: A Zero-Shot NAS for High-Performance Deep Image Recognition](https://mlsanity.com/paper/2102.01063/) (2021)
- [Speedy Performance Estimation for Neural Architecture Search](https://mlsanity.com/paper/2006.04492/) (2020)
- [Optuna vs Code Llama: Are LLMs a New Paradigm for Hyperparameter Tuning?](https://mlsanity.com/paper/2504.06006/) (2025)
- [LEMUR Neural Network Dataset: Towards Seamless AutoML](https://mlsanity.com/paper/2504.10552/) (2025)
- [NNGPT: Rethinking AutoML with Large Language Models](https://mlsanity.com/paper/2511.20333/) (2025)
- [VIST-GPT: Ushering in the Era of Visual Storytelling with LLMs?](https://mlsanity.com/paper/2504.19267/) (2025)
- [Preparation of Fractal-Inspired Computational Architectures for Advanced Large Language Model Analysis](https://mlsanity.com/paper/2511.07329/) (2025)
- [A Retrieval-Augmented Generation Approach to Extracting Algorithmic Logic from Neural Networks](https://mlsanity.com/paper/2512.04329/) (2025)
- [AugmentGest: Can Random Data Cropping Augmentation Boost Gesture Recognition Performance?](https://mlsanity.com/paper/2506.07216/) (2025)
- [Enhancing LLM-Based Neural Network Generation: Few-Shot Prompting and Efficient Validation for Automated Architecture Design](https://mlsanity.com/paper/2512.24120/) (2025)
- [LLM as a Neural Architect: Controlled Generation of Image Captioning Models Under Strict API Contracts](https://mlsanity.com/paper/2512.14706/) (2025)
- [SEKI: Self-Evolution and Knowledge Inspiration based Neural Architecture Search via Large Language Models](https://mlsanity.com/paper/2502.20422/) (2025)
- [The Unreasonable Effectiveness Of Early Discarding After One Epoch In Neural Network Hyperparameter Optimization](https://mlsanity.com/paper/2404.04111/) (2024)

## Built on by (most cited first, 10 of 10)

- [From Brute Force to Semantic Insight: Performance-Guided Data Transformation Design with LLMs](https://mlsanity.com/paper/2601.03808/) (2026)
- [Resource-Efficient Iterative LLM-Based NAS with Feedback Memory](https://mlsanity.com/paper/2603.12091/) (2026)
- [Closed-Loop LLM Discovery of Non-Standard Channel Priors in Vision Models](https://mlsanity.com/paper/2601.08517/) (2026)
- [Delta-Based Neural Architecture Search: LLM Fine-Tuning via Code Diffs](https://mlsanity.com/paper/2605.04903/) (2026)
- [LEMUR 2: Unlocking Neural Network Diversity for AI](https://mlsanity.com/paper/2607.06839/) (2026)
- [Convergence Theory for Iterative LLM-Based Neural Architecture Search: A Parametric Cross-Entropy Framework with Closed-Form Proxy Reliability](https://mlsanity.com/paper/2605.30103/) (2026)
- [Systematic Exploration of 4-Expert Heterogeneous Mixture-of-Experts via Automated Pipeline Search](https://mlsanity.com/paper/2606.23739/) (2026)
- [From Code to Prediction: Fine-Tuning LLMs for Neural Network Performance Classification in NNGPT](https://mlsanity.com/paper/2605.03686/) (2026)
- [Towards Robust Training in NNGPT AutoML Pipeline: A Loss-Optimizer Pairing Selection Study](https://mlsanity.com/paper/2606.20933/) (2026)
- [Systematic Evaluation of Learning Rate Scheduling Strategies Across Heterogeneous Architectures](https://mlsanity.com/paper/2607.08511/) (2026)

## Abstract (arXiv)

Large language models (LLMs) excel in program synthesis, yet their capacity for neural architecture design -- balancing syntactic reliability, performance, and structural novelty -- remains underexplored. We present a closed-loop architecture synthesis pipeline within the NNGPT framework, in which a code-oriented LLM evolves over 22 supervised fine-tuning cycles. At each cycle, the LLM synthesizes PyTorch convolutional networks, validated via low-fidelity performance signals and filtered via a MinHash--Jaccard criterion to prevent structural redundancy before being incorporated into the LEMUR dataset. High-performing candidates with novel architectures are converted into prompt--code pairs for parameter-efficient LoRA fine-tuning. This feedback loop drives a measurable distributional shift, progressively internalizing empirical architectural priors such that valid and high-performing outputs evolve from scarce to dominant across cycles. On CIFAR-10, the valid generation rate stabilizes at 50.6% (peaking at 74.5%), mean first-epoch accuracy rises from 28.1% to 51.0%, and candidates exceeding 40% accuracy grow from 2.0% to 96.8%. Cross-dataset transfer to CIFAR-100 and SVHN confirms that improved validity, shifted accuracy distributions, and sustained novelty generalize across benchmarks of varying difficulty and visual domain. Across 22 cycles, 455 unique architectures absent from the original corpus are admitted under the novelty filter. By grounding synthesis in execution feedback and novelty filtering, we demonstrate that iterative self-supervised fine-tuning reshapes an LLM into a task-specialized architectural prior -- improving generation reliability, proxy performance, and structural diversity -- offering a reproducible, annotation-free alternative to hand-crafted search spaces.