analiza aiqbench
This research introduces a geometric filtering framework that uses Euclidean distance in embedding space to validate synthetic data generated by LLMs, significantly improving few-shot text classification performance over standard augmentation techniques like SMOTE. The study demonstrates that simple distance-based metrics outperform complex multi-criteria filters across diverse datasets and models without requiring specific modifications for different tasks such as named entity recognition.
Geometric Filtering of LLM-Generated Samples for Few-Shot Text Classification
arXiv:2608.13866v1 Announce Type: cross Abstract: Large language models (LLMs) can generate synthetic training data for text classification, but the quality of generated samples is heterogeneous: some fall in correct class regions of the embedding space while others land in peripheral or cross-class zones. We propose a geometric filtering framework that evaluates each LLM-generated sample by its Euclidean distance to real class examples in a sentence embedding space, selecting only geometrically consistent candidates. A soft weighting mechanism transforms filter scores into sample weights for classifier training. Evaluated across 13 datasets, 5 classifiers, 10 augmentation methods, and over 6,700 configurations, our method achieves +2.61 percentage points (pp) over SMOTE ($p<0.0001$, Cohen's $d=0.95$, 88.9% win rate). The approach generalizes to named entity recognition (+9.26pp, 100% win rate) without filter modification, and is robust across 5 LLMs from 4 providers. A key finding is that the simplest distance-based filter consistently outperforms complex multi-criteria alternatives.
Abstract:Large language models (LLMs) can generate synthetic training data for text classification, but the quality of generated samples is heterogeneous: some fall in correct class regions of the embedding space while others land in peripheral or cross-class zones. We propose a geometric filtering framework that evaluates each LLM-generated sample by its Euclidean distance to real class examples in a sentence embedding space, selecting only geometrically consistent candidates. A soft weighting mechanism transforms filter scores into sample weights for classifier training. Evaluated across 13 datasets, 5 classifiers, 10 augmentation methods, and over 6,700 configurations, our method achieves +2.61 percentage points (pp) over SMOTE ($p<0.0001$, Cohen's $d=0.95$, 88.9% win rate). The approach generalizes to named entity recognition (+9.26pp, 100% win rate) without filter modification, and is robust across 5 LLMs from 4 providers. A key finding is that the simplest distance-based filter consistently outperforms complex multi-criteria alternatives.
Machine Learning (cs.LG); Computation and Language (cs.CL)
arXiv:2608.13866 [cs.LG]
(or arXiv:2608.13866v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.13866 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 243 cuvinte
Subiecte conexe
PRX Part 4: Our Data Strategy
Hugging Face · 6 iul. 2026
Exact Degeneracy Under Balanced k-Shot Sampling:Consequences for Small-Sample Discriminant Analysis on LLM Embeddings
arXiv · 10 sept. 2026
Task-Seeded Synthetic Q&A Generation for Nemotron Pretraining
Hugging Face · 4 iun. 2026
Open Distillation of Hereditary Traits
AI Alignment Forum · 14 iul. 2026