# Topic Over Source: The Key to Effective Data Mixing for Language Models Pre-training

arXiv: https://arxiv.org/abs/2502.16802 · HTML: https://mlsanity.com/paper/2502.16802/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 29 references that are in the map.
- First posted on arXiv 24 Feb 2025.

## Builds on (most cited first, 19 of 19)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [PIQA: Reasoning about Physical Commonsense in Natural Language](https://mlsanity.com/paper/1911.11641/) (2019)
- [Billion-scale similarity search with GPUs](https://mlsanity.com/paper/1702.08734/) (2017)
- [Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering](https://mlsanity.com/paper/1809.02789/) (2018)
- [Scaling Language Models: Methods, Analysis & Insights from Training Gopher](https://mlsanity.com/paper/2112.11446/) (2021)
- [The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale](https://mlsanity.com/paper/2406.17557/) (2024)
- [RACE: Large-scale ReAding Comprehension Dataset From Examinations](https://mlsanity.com/paper/1704.04683/) (2017)
- [A Pretrainer's Guide to Training Data: Measuring the Effects of Data Age, Domain Coverage, Quality, & Toxicity](https://mlsanity.com/paper/2305.13169/) (2023)
- [RegMix: Data Mixture as Regression for Language Model Pre-training](https://mlsanity.com/paper/2407.01492/) (2024)
- [DoGE: Domain Reweighting with Generalization Estimation](https://mlsanity.com/paper/2310.15393/) (2023)
- [Organize the Web: Constructing Domains Enhances Pre-Training Data Curation](https://mlsanity.com/paper/2502.10341/) (2025)
- [MLlib: Machine Learning in Apache Spark](https://mlsanity.com/paper/1505.06807/) (2015)
- [Large Language Models Offer an Alternative to the Traditional Approach of Topic Modelling](https://mlsanity.com/paper/2403.16248/) (2024)
- [Data, Data Everywhere: A Guide for Pretraining Dataset Construction](https://mlsanity.com/paper/2407.06380/) (2024)
- [Addressing Topic Granularity and Hallucination in Large Language Models for Topic Modelling](https://mlsanity.com/paper/2405.00611/) (2024)

## Built on by (most cited first, 4 of 4)

- [GEM: Geometric Entropy Mixing for Optimal LLM Data Curation](https://mlsanity.com/paper/2605.26121/) (2026)
- [topicwizard -- a Modern, Model-agnostic Framework for Topic Model Visualization and Interpretation](https://mlsanity.com/paper/2505.13034/) (2025)
- [HERMES: A Multi-Granularity Labeling Substrate for Pre-training Data Mixtures](https://mlsanity.com/paper/2607.02266/) (2026)
- [Curriculum Learning as Transport: Understanding Curricula with Wasserstein Geodesics](https://mlsanity.com/paper/2609.09099/) (2026)

## Abstract (arXiv)

The performance of large language models (LLMs) is significantly affected by the quality and composition of their pre-training data, which is inherently diverse, spanning various languages, sources, and topics. Effectively integrating these heterogeneous data groups is crucial for optimizing LLM performance. Previous research has predominantly concentrated on source-based data mixing, often neglecting the nuanced topic-level characteristics of the data. To address this gap, we propose a topic-based data mixing strategy that utilizes detailed topic labels generated through a multi-stage process combining unsupervised clustering, LLM-based summarization, and supervised classifier training. With this strategy, we conduct the first comprehensive comparison of topic-based versus source-based partitioning across multiple mixing strategies. We demonstrate that language models pretrained on data mixed by topics consistently outperform those trained on data mixed by sources across multiple methods including RegMix, DoReMi,temperature-based sampling, and a manual mixing method based on downstream task performance. Our theoretical analysis reveals that topic-based data achieves significantly lower validation loss compared to source-based approaches, creating a better optimization landscape for model training. We will make our code, annotated datasets, and topic classification models publicly available to facilitate further research.