Neural architecture search (NAS) traditionally requires significant human expertise or automated trial-and-error to design deep learning models. We present NN-Caption, an LLM-guided neural architecture search pipeline that generates runnable image-captioning models by composing CNN encoders from LEMUR’s classification backbones with sequence decoders (LSTM/GRU/Transformer) under a strict Net API [1,2]. Using DeepSeek-R1-0528-Qwen3-8B as the primary generator [3], we present the prompt template and examples of generated architectures. We evaluate on MS COCO with BLEU-4 [4,5]. The LLM generated dozens of captioning models, with over half successfully trained and producing meaningful captions. We analyse the outcomes of using different numbers of input model snippets (5 vs. 10) in the prompt, finding a slight drop in success rate when providing more candidate components. We also report training dynamics (caption accuracy vs. epochs) and the highest BLEU-4 attained. Our results highlight the promise of LLM-guided NAS: the LLM not only proposes architectures but also suggests hyperparameters and training practices. We identify the challenges encountered (e.g., code hallucinations or API compliance issues) and detail how prompt rules and iterative code fixes addressed them. This work presents a pipeline that integrates prompt-based code generation with automatic evaluation, and adds dozens of novel captioning models to the open LEMUR dataset to facilitate reproducible benchmarking and downstream AutoML research.
We present ANYU, a new virtually augmented version of the NYU depth v2 dataset, designed for monocular depth estimation. In contrast to the well-known approach where full 3D scenes of a virtual world are utilized to generate artificial datasets, ANYU was created by incorporating RGB-D representations of virtual reality objects into the original NYU depth v2 images. We specifically did not match each generated virtual object with an appropriate texture and a suitable location within the real-world image. Instead, an assignment of texture, location, lighting, and other rendering parameters was randomized to maximize a diversity of the training data, and to show that it is randomness that can improve the generalizing ability of a dataset. By conducting extensive experiments with our virtually modified dataset and validating on the original NYU depth v2 and iBims-1 benchmarks, we show that ANYU improves the monocular depth estimation performance and generalization of deep neural networks with considerably different architectures, especially for the current state-of-the-art VPD model. To the best of our knowledge, this is the first work that augments a real-world dataset with randomly generated virtual 3D objects for monocular depth estimation. We make our ANYU dataset publicly available in two training configurations with 10% and 100% additional synthetically enriched RGB-D pairs of training images, respectively, for efficient training and empirical exploration of virtual augmentation at https://github.com/ABrain-One/ANYU
This paper defines a task that utilizes vision and language models to improve benchmarks through analysis of CIFAR-10 and CIFAR-100 datasets. The work divides its operations into image categorization followed by visual description production. The task utilizes BEiT and Swin models as state-of-the-art application-specific components for both parts of this research. We selected the current best image classification checkpoints available in the market which delivered 99.00% accuracy on CIFAR-10 and 92.01% on CIFAR-100. For dense contextually rich text output we used BLIP. The expert models performed well on their target responsibilities using minimal noisy data. The BART model achieved new state-of-the-art accuracies when used as a text classifier to compare synthesized descriptions and reached 99.73% accuracy on CIFAR-10 while attaining 98.38% accuracy on CIFAR-100. This paper demonstrates how our integrated vision and language decomposition-hierarchical model surpasses all existing state-of-the-art results on these common benchmark classifications. The full framework, along with the classified images and generated datasets, is available at https://github.com/bhavyarupani/LLM-Img-Classification.
Deploying deep vision models on edge hardware requires understanding how architecture and training hyperparameters jointly determine accuracy and inference latency, yet these relationships remain poorly characterized in a systematic, data-driven manner. This paper presents a two-stage statistical framework providing interpretable, closed-form insights into both. In the first stage, we apply distance correlation (dCor) and the maximal information coefficient (MIC) across seven image-classification datasets, revealing that batch size and total layer count are the strongest universal accuracy predictors (mean dCor: 0.228 and 0.174), while learning rate achieves the highest MIC (0.226), reflecting a non-monotonic relationship with accuracy. In the second stage, PySR symbolic regression (representing, to our knowledge, the first application to cross-dataset vision model accuracy prediction) derives compact, interpretable formulas. Dataset-specific models achieve from 0.20 to 0.45; a universal model achieves a mean leave-one-dataset-out of 0.23, remaining strictly positive on all held-out datasets, whereas ordinary linear regression collapses to . We further derive device-specific inference latency formulas for CPU, GPU, and NPU, outperforming classical baselines by – in and confirming fundamental device heterogeneity. Together, these results offer interpretable surrogate models for screening deep vision architectures under accuracy and latency constraints in edge deployment.
This work presents the integration of Mixture of Experts (MoE) architectures into the LEMUR neural network dataset to enhance model diversity and scalability. The MoE framework employs multiple expert networks and a gating mechanism for dynamic routing, enabling efficient computation and improved specialization across tasks. Eight MoE variants were implemented and benchmarked on CIFAR-10, achieving up to 93% accuracy with optimized routing, regularization, and training strategies. This integration provides a foundation for benchmarking expert-based models within LEMUR and supports future research in adaptive model composition and automated machine learning. The project work and its plugins are accessible as open source projects under the MIT license at https://github.com/ABrain-One/nn-dataset.