cs.LGMar 16, 2026

Data-Local Autonomous LLM-Guided Neural Architecture Search for Multiclass Multimodal Time-Series Classification

Authors: Emil Hardarson, Luka Biedebach, Ómar Bessi Ómarsson, Teitur Hrólfsson, Anna Sigridur Islind, María Óskarsdóttir

Organizations: Reykjavik University, Reykjavik, Iceland · University of Southampton, Southampton, United Kingdom

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Applying machine learning to sensitive time-series data is often bottlenecked by the iteration loop: Performance depends strongly on preprocessing and architecture, yet training often has to run on-premise under strict data-local constraints. This is a common problem in healthcare and other privacy-constrained domains (e.g., a hospital developing deep learning models on patient EEG). This bottleneck is particularly challenging in multimodal fusion, where sensor modalities must be individually preprocessed and then combined. LLM-guided neural architecture search (NAS) can automate this exploration, but most existing workflows assume cloud execution or access to data-derived artifacts that cannot be exposed. We present a novel data-local, LLM-guided search framework that handles candidate pipelines remotely while executing all training and evaluation locally under a fixed protocol. The controller observes only trial-level summaries, such as pipeline descriptors, metrics, learning-curve statistics, and failure logs, without ever accessing raw samples or intermediate feature representations. Our framework targets multiclass, multimodal learning via one-vs-rest binary experts per class and modality, a lightweight fusion MLP, and joint search over expert architectures and modality-specific preprocessing. We evaluate our method on two regimes: UEA30 (public multivariate time-series classification dataset) and SleepEDFx sleep staging (heterogeneous clinical modalities such as EEG, EOG, and EMG). The results show that the modular baseline model is strong, and the LLM-guided NAS further improves it. Notably, our method finds models that perform within published ranges across most benchmark datasets. Across both settings, our method reduces manual intervention by enabling unattended architecture search while keeping sensitive data on-premise.

Figures & tables

Figure 1: Overview of the expert decomposition and representation-level fusion.
Figure 2: Schematic of the data-local autonomous NAS loop. The data and all training remain on-premise. The local execution (model debugging, training and hyperparameter optimization) consumes directives and writes results to a shared filesystem. A remote LLM interacts only through these non-sensitive files. A smaller, local LLM debugs simple errors in the code proposed by the remote LLM, since error messages may contain sensitive data.
DatasetMod, mmDim, dmd_{m}Len, TmT_{m}Class, CCTrain sizeTest sizeBaseline end-to-endBaseline stagedLLM NASPrevious work
SEDF3530005112,28710,77482.984.787.9 a[82.8 - 94.0]
AWR191442527530091.796.3– b[94.3 - 99.6]
AF126403151533.333.3–[6.7 - 74.0]
BM261004404010092.5– b[72.2 - 100.0]
CT130201422143690.195.895.8[98.7 - 99.6]
CR161197121087295.893.1– b[93.1 - 100.0]
Table 1: Benchmark overview and performance summary. For each dataset, the table reports the number of modalities ( mm ), number of variates ( dmd_{m} ), sequence length ( TmT_{m} ), number of classes ( CC ), train and test set sizes, and test accuracy (%) for three configurations: the end-to-end baseline, the staged baseline with individually trained binary experts followed by fusion, and the proposed LLM-guided NAS model. The final column reports the range of accuracies reported in recent published work for the respective dataset. Unless otherwise stated, NAS was run with a budget of 10 cycles. “–” indicates that the configuration was not evaluated, either because the dataset was too small or because the baseline model achieved 100% accuracy on the validation set. 10 cycle searches unless stated otherwise.
Figure 3: Search trajectories for binary experts during a 30-cycle NAS on the SleepEDFx dataset. Each panel corresponds to one class-modality expert. Dashed lines show the F 1 score of the best found expert, and circles show the F 1 score of the expert trained during the cycle.
Figure 4: Search trajectories for binary experts during a 60-cycle NAS on the EthanolConcentration dataset. Each panel corresponds to one class-modality expert. Dashed lines show the F 1 score of the best found expert, and circles show the F 1 score of the expert trained during the cycle.
More in this topic

Large Language Model-Guided NAS

Large language model-guided neural architecture search uses language models to generate and optimize neural network designs.

See all 24 papers

Explore similar work

CardsList
Jul 8, 2026

Agentic Neural Architecture Search

Neural architecture search (NAS) methods have grown increasingly efficient, yet they remain bounded by manually engineered search spaces that require substantial domain expertise and must be rebuilt for every new task. Large language models (LLMs) can generate architectures in an open-ended space, but how to optimally divide the labor between LLM-driven design and NAS-driven search remains unexplored. We propose a mechanism that bridges these two paradigms: an LLM produces a high-quality seed architecture, then decomposes it into a "slotted architecture", a scaffold with named, interchangeable module slots that automatically defines a bounded, task-specific search space for conventional NAS to explore, without manual engineering. We instantiate this mechanism in AgentNAS, a modular three-phase pipeline in which each component's contribution can be measured independently. On 17 tasks spanning classification, dense regression, segmentation, and multi-label tagging across diverse modalities (NAS-Bench-360 and Unseen NAS), AgentNAS establishes a new state of the art on 11 tasks, outperforming published baselines including task-specific expert designs. Ablation studies show that the two search mechanisms are broadly complementary: the LLM-generated seed already surpasses published baselines on the majority of tasks, and NAS delivers additional gains in most cases through combinatorial recombination across slots, a mode of search that independent LLM samples cannot replicate. These patterns hold across three LLMs of different capability levels, confirming that the division of labor is robust. Our code is available at https://github.com/alroimfebruary/AgentNAS.

Seokhoon Jeong, Mijung Kim, Taehwan Kim

Artificial Intelligence Graduate School, Ulsan National Institute of Science and Engineering, Ulsan, Republic of Korea · Department of Computer Science and Engineering, Ulsan National Institute of Science and Engineering, Ulsan, Republic of Korea

Large Language Model-Guided NASLarge Language Model-Guided OptimizationNeural Architecture Search
AskPDF
Mar 12, 2026

Resource-Efficient Iterative LLM-Based NAS with Feedback Memory

Neural Architecture Search (NAS) automates network design, but conventional methods demand substantial computational resources. We propose a closed-loop pipeline leveraging large language models (LLMs) to iteratively generate, evaluate, and refine convolutional neural network architectures for image classification on a single consumer-grade GPU without LLM fine-tuning. Central to our approach is a historical feedback memory inspired by Markov chains: a sliding window of K=5K{=}5 recent improvement attempts keeps context size constant while providing sufficient signal for iterative learning. Unlike prior LLM optimizers that discard failure trajectories, each history entry is a structured diagnostic triple -- recording the identified problem, suggested modification, and resulting outcome -- treating code execution failures as first-class learning signals. A dual-LLM specialization reduces per-call cognitive load: a Code Generator produces executable PyTorch architectures while a Prompt Improver handles diagnostic reasoning. Since both the LLM and architecture training share limited VRAM, the search implicitly favors compact, hardware-efficient models suited to edge deployment. We evaluate three frozen instruction-tuned LLMs (≤7{\leq}7B parameters) across up to 2000 iterations in an unconstrained open code space, using one-epoch proxy accuracy on CIFAR-10, CIFAR-100, and ImageNette as a fast ranking signal. On CIFAR-10, DeepSeek-Coder-6.7B improves from 28.2% to 69.2%, Qwen2.5-7B from 50.0% to 71.5%, and GLM-5 from 43.2% to 62.0%. A full 2000-iteration search completes in ≈18{\approx}18 GPU hours on a single RTX~4090, establishing a low-budget, reproducible, and hardware-aware paradigm for LLM-driven NAS without cloud infrastructure.

Xiaojie Gu, Dmitry Ignatov, Radu Timofte

Computer Vision Lab, CAIDAS & IFI, University of Würzburg, Germany

Large Language Model-Guided NASNeural Architecture SearchLarge Language Model-Guided OptimizationHardware-Aware NAS
AskPDF
Apr 17, 2026

LLM as a Tool, Not an Agent: Code-Mined Tree Transformations for Neural Architecture Search

Neural Architecture Search (NAS) aims to automatically discover high-performing deep neural network (DNN) architectures. However, conventional algorithm-driven NAS relies on carefully hand-crafted search spaces to ensure executability, which restricts open-ended exploration. Recent coding-based agentic approaches using large language models (LLMs) reduce manual design, but current LLMs struggle to reliably generate complex, valid architectures, and their proposals are often biased toward a narrow set of patterns observed in their training data. To bridge reliable algorithmic search with powerful LLM assistance, we propose LLMasTool, a hierarchical tree-based NAS framework for stable and open-ended model evolution. Our method automatically extracts reusable modules from arbitrary source code and represents full architectures as hierarchical trees, enabling evolution through reliable tree transformations rather than code generation. At each evolution step, coarse-level planning is governed by a diversity-guided algorithm that leverages Bayesian modeling to improve exploration efficiency, while the LLM resolves the remaining degrees of freedom to ensure a meaningful evolutionary trajectory and an executable generated architecture. With this formulation, instead of fully agentic LLM approaches, our method explores diverse directions beyond the inherent biases in the LLM. Our method improves over existing NAS methods by 0.69, 1.83, and 2.68 points on CIFAR-10, CIFAR-100, and ImageNet16-120, demonstrating its effectiveness.

Masakazu Yoshimura, Zitang Sun, Yuiko Sakuma +3

Sony Group Corporation, Tokyo, Japan

Large Language Model-Guided NASLarge Language Model-Guided OptimizationNeural Architecture Search
AskPDF