UniGeM: Unifying Data Mixing and Selection via Geometric Exploration and Mining

TL;DR

UniGeM is a novel framework that unifies data mixing and selection for large language models by using geometric exploration, leading to improved data efficiency and model performance without external datasets.

Contribution

UniGeM introduces a hierarchical, geometry-based approach to unify data mixing and selection, eliminating the need for proxy models or external references.

Findings

Achieves 2.0× data efficiency over random baseline

Improves reasoning and multilingual performance

Validates on 8B and 16B MoE models with 100B tokens

Abstract

The scaling of Large Language Models (LLMs) is increasingly limited by data quality. Most methods handle data mixing and sample selection separately, which can break the structure in code corpora. We introduce \textbf{UniGeM}, a framework that unifies mixing and selection by treating data curation as a \textit{manifold approximation} problem without training proxy models or relying on external reference datasets. UniGeM operates hierarchically: \textbf{Macro-Exploration} learns mixing weights with stability-based clustering; \textbf{Micro-Mining} filters high-quality instances by their geometric distribution to ensure logical consistency. Validated by training 8B and 16B MoE models on 100B tokens, UniGeM achieves \textbf{2.0$\times$ data efficiency} over a random baseline and further improves overall performance compared to SOTA methods in reasoning-heavy evaluations and multilingual generalization.