# LLM-Driven Neural Network Generation with Same-Family Architecture Guidance: Disentangling Transfer and Adaptation

Source: [LLM-Driven Neural Network Generation with Same-Family Architecture Guidance: Disentangling Transfer and Adaptation](<https://www.sota2.com/research/paper/6a4dd9b24da6617fe9198c3f>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-07-09T05:00:01.225Z

Authors: Kabir Dev Paul Baghel, Radu Timofte, Dmitry Ignatov

Published: 2026-07-06T23:59:18.000Z

## Abstract

Large language models (LLMs) can generate neural-network modifications, but unrestricted generation is often invalid or harmful. This paper studies a narrower setting: improving a weak target model using a stronger same-family source model from a neural-network database. We propose a source-guided candidate-generation protocol with non-source controls, source-conditioned candidates, and a no-LLM hp\_copy ablation under equal evaluation budgets. The protocol reports validity separately from accuracy and selects the best valid candidate only when it improves the target. On CIFAR-10, the strongest source-guided candidate reaches 0.5049 accuracy versus 0.2398 for the best non-source candidate, a +0.2651 advantage, while improving a weak target originally at 0.1254; a five-epoch check preserves the gain at 0.7686 versus 0.4839. On SVHN AlexNet with DeepSeek-Coder-6.7B, source-guided transfer reaches 0.7880 versus 0.2254, a +0.5626 advantage; a fresh repeat reaches 0.8069 versus 0.2509, a +0.5560 advantage. Direct source-recipe copy produces 0.1959 on SVHN AlexNet, matching the original target, while hp\_transfer reaches 0.7880, showing that the LLM adapts rather than copies the source recipe. Family-level analysis shows the clearest positive signals for AlexNet, with 6/8 wins across SVHN, Imagenette, and CelebA-Gender, and alt\_nn1, with 8/10 wins on CIFAR-10.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Image Classification on CIFAR-10 (test) (Validity and Accuracy Advantage)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-validity-and-accuracy-advantage.md>): Image Classification; CIFAR-10 (test); Validity Count: 64.

- [Hyperparameter Search (AlexNet) on SVHN](<https://www.sota2.com/research/sota/hyperparameter-search-alexnet-on-svhn.md>): Hyperparameter Search (AlexNet); SVHN; Non-source Validity Count: not recorded.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2607.05704>)

[PDF](<https://arxiv.org/pdf/2607.05704.pdf>)