Optimizing Pretraining Data Mixtures with LLM-Estimated Utility

• URL: http://arxiv.org/abs/2501.11747v2
• Date: Thu, 23 Jan 2025 20:45:47 GMT
• Title: Optimizing Pretraining Data Mixtures with LLM-Estimated Utility
• Authors: William Held, Bhargavi Paranjape, Punit Singh Koura, Mike Lewis, Frank Zhang, Todor Mihaylov,
• Abstract summary: Large Language Models improve with increasing amounts of high-quality training data. We find token-counts outperform manual and learned mixes, indicating that simple approaches for dataset size and diversity are surprisingly effective. We propose two complementary approaches: UtiliMax, which extends token-based $200s by incorporating utility estimates from reduced-scale ablations, achieving up to a 10.6x speedup over manual baselines; and Model Estimated Data Utility (MEDU), which leverages LLMs to estimate data utility from small samples, matching ablation-based performance while reducing computational requirements by $simx.
• Score: 52.08428597962423
• License:
• Abstract: Large Language Models improve with increasing amounts of high-quality training data. However, leveraging larger datasets requires balancing quality, quantity, and diversity across sources. After evaluating nine baseline methods under both compute- and data-constrained scenarios, we find token-count heuristics outperform manual and learned mixes, indicating that simple approaches accounting for dataset size and diversity are surprisingly effective. Building on this insight, we propose two complementary approaches: UtiliMax, which extends token-based heuristics by incorporating utility estimates from reduced-scale ablations, achieving up to a 10.6x speedup over manual baselines; and Model Estimated Data Utility (MEDU), which leverages LLMs to estimate data utility from small samples, matching ablation-based performance while reducing computational requirements by $\sim$200x. Together, these approaches establish a new framework for automated, compute-efficient data mixing that is robust across training regimes.

Related papers

• Star-Agents: Automatic Data Optimization with LLM Agents for Instruction Tuning [71.2981957820888]
  We propose a novel Star-Agents framework, which automates the enhancement of data quality across datasets. The framework initially generates diverse instruction data with multiple LLM agents through a bespoke sampling method. The generated data undergo a rigorous evaluation using a dual-model method that assesses both difficulty and quality.
  arXiv  Detail & Related papers  (2024-11-21T02:30:53Z)
• Efficient Alignment of Large Language Models via Data Sampling [0.4915744683251149]
  We propose an information theory-based methodology for efficient alignment by identifying a small high quality subset. We find that the model aligned using our proposed methodology outperforms other sampling methods and performs comparable to the model aligned with the full dataset.
  arXiv  Detail & Related papers  (2024-11-15T19:36:15Z)
• AutoScale: Automatic Prediction of Compute-optimal Data Composition for Training LLMs [61.13296177652599]
  This paper demonstrates that the optimal composition of training data from different domains is scale-dependent. We introduce *AutoScale*, a novel, practical approach for optimizing data compositions at potentially large training data scales. Our evaluation on GPT-2 Large and BERT pre-training demonstrates *AutoScale*'s effectiveness in improving training convergence and downstream performance.
  arXiv  Detail & Related papers  (2024-07-29T17:06:30Z)
• SoftDedup: an Efficient Data Reweighting Method for Speeding Up Language Model Pre-training [12.745160748376794]
  We propose a soft deduplication method that maintains dataset integrity while selectively reducing the sampling weight of data with high commonness. Central to our approach is the concept of "data commonness", a metric we introduce to quantify the degree of duplication. Empirical analysis shows that this method significantly improves training efficiency, achieving comparable perplexity scores with at least a 26% reduction in required training steps.
  arXiv  Detail & Related papers  (2024-07-09T08:26:39Z)
• BiMix: A Bivariate Data Mixing Law for Language Model Pretraining [47.77701041534746]
  The impact of pretraining data composition on model performance remains poorly understood. $textbfBiMix$ provides a systematic framework for understanding and optimizing data mixtures. Our work contributes both theoretical insights into data mixing dynamics and practical tools for enhancing LLM training efficiency.
  arXiv  Detail & Related papers  (2024-05-23T09:44:02Z)
• Data Mixing Laws: Optimizing Data Mixtures by Predicting Language Modeling Performance [55.872926690722714]
  We study the predictability of model performance regarding the mixture proportions in function forms. We propose nested use of the scaling laws of training steps, model sizes, and our data mixing law. Our method effectively optimize the training mixture of a 1B model trained for 100B tokens in RedPajama.
  arXiv  Detail & Related papers  (2024-03-25T17:14:00Z)
• How to Train Data-Efficient LLMs [56.41105687693619]
  We study data-efficient approaches for pre-training language models (LLMs) We find that Ask-LLM and Density sampling are the best methods in their respective categories. In our comparison of 19 samplers, involving hundreds of evaluation tasks and pre-training runs, we find that Ask-LLM and Density are the best methods in their respective categories.
  arXiv  Detail & Related papers  (2024-02-15T02:27:57Z)
• Deep Learning with Multiple Data Set: A Weighted Goal Programming Approach [2.7393821783237184]
  Large-scale data analysis is growing at an exponential rate as data proliferates in our societies. Deep Learning models require plenty of resources, and distributed training is needed. This paper presents a Multicriteria approach for distributed learning.
  arXiv  Detail & Related papers  (2021-11-27T07:10:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.