# LLM-Driven Performance-Space Augmentation for Meta-Learning-Based Algorithm Selection

arXiv: https://arxiv.org/abs/2605.09518 · HTML: https://mlsanity.com/paper/2605.09518/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 49 references that are in the map.
- First posted on arXiv 10 May 2026.

## Builds on (most cited first, 9 of 9)

- [Optuna: A Next-generation Hyperparameter Optimization Framework](https://mlsanity.com/paper/1907.10902/) (2019)
- [Language Models are Realistic Tabular Data Generators](https://mlsanity.com/paper/2210.06280/) (2022)
- [Curated LLM: Synergy of LLMs and Data Curation for tabular augmentation in low-data regimes](https://mlsanity.com/paper/2312.12112/) (2023)
- [EPIC: Effective Prompting for Imbalanced-Class Data Synthesis in Tabular Data Classification via Large Language Models](https://mlsanity.com/paper/2404.12404/) (2024)
- [TabGen-ICL: Residual-Aware In-Context Example Selection for Tabular Data Generation](https://mlsanity.com/paper/2502.16414/) (2025)
- [AIGT: AI Generative Table Based on Prompt](https://mlsanity.com/paper/2412.18111/) (2024)
- [A Feature Subset Selection Algorithm Automatic Recommendation Method](https://mlsanity.com/paper/1402.0570/) (2014)
- [Distances between Data Sets Based on Summary Statistics](https://mlsanity.com/paper/1902.01334/) (2019)
- [Transforming Datasets to Requested Complexity with Projection-based Many-Objective Genetic Algorithm](https://mlsanity.com/paper/2507.15132/) (2025)

## Abstract (arXiv)

Meta-learning for algorithm selection relies on a meta-dataset in which each row corresponds to a supervised learning dataset described by meta-features and labelled with a target value that is associated with algorithm choice (typically, some function of algorithm performance). A persistent limitation is that the number of curated real-world datasets is small, resulting in sparse meta-datasets that constrain meta-learner generalisation. In this paper, we address this problem by augmenting the meta-dataset with synthetic regression datasets produced via a large language model (LLM), with generation steered toward target regions of a low-dimensionality performance space. In our experiments, we adopt a two-dimensional geometric setting defined by the cross-validated R2 scores of two anchor algorithms, known as landmarkers. We compare two augmentation strategies: (1) uniform sampling, which distributes synthetic datasets across the performance space; and (2) margin-based sampling, which concentrates them near the decision boundary where landmarker preference is most ambiguous. Across 42 real-world UCI regression datasets and 730 synthetic datasets, both strategies substantially improve meta-learner performance over the unaugmented baseline under regression and multi-label evaluation formulations. However, uniform augmentation consistently outperforms margin-based augmentation, achieving a 17.47% relative reduction in Hamming loss, a 100.41% relative improvement in subset accuracy, and a +6.09% relative gain in pooled out-of-fold R2. These results lead us to postulate a central thesis: the performance of algorithms resides on a low-dimensional performance manifold, whose reconstruction bias may be minimised by user-guided LLMs that seek to maximise uniform ε-cover, and consequently, lead to improved meta-learning for algorithm selection.