Entropy Law: The Story Behind Data Compression and LLM Performance
- URL: http://arxiv.org/abs/2407.06645v3
- Date: Thu, 11 Jul 2024 03:06:45 GMT
- Title: Entropy Law: The Story Behind Data Compression and LLM Performance
- Authors: Mingjia Yin, Chuhan Wu, Yufei Wang, Hao Wang, Wei Guo, Yasheng Wang, Yong Liu, Ruiming Tang, Defu Lian, Enhong Chen,
- Abstract summary: We find that model performance is negatively correlated to the compression ratio of training data, which usually yields a lower training loss.
Based on the findings of the entropy law, we propose a quite efficient and universal data selection method.
We also present an interesting application of entropy law that can detect potential performance risks at the beginning of model training.
- Score: 115.70395740286422
- License: http://creativecommons.org/licenses/by/4.0/
- Abstract: Data is the cornerstone of large language models (LLMs), but not all data is useful for model learning. Carefully selected data can better elicit the capabilities of LLMs with much less computational overhead. Most methods concentrate on evaluating the quality of individual samples in data selection, while the combinatorial effects among samples are neglected. Even if each sample is of perfect quality, their combinations may be suboptimal in teaching LLMs due to their intrinsic homogeneity or contradiction. In this paper, we aim to uncover the underlying relationships between LLM performance and data selection. Inspired by the information compression nature of LLMs, we uncover an ``entropy law'' that connects LLM performance with data compression ratio and first-epoch training loss, which reflect the information redundancy of a dataset and the mastery of inherent knowledge encoded in this dataset, respectively. Through both theoretical deduction and empirical evaluation, we find that model performance is negatively correlated to the compression ratio of training data, which usually yields a lower training loss. Based on the findings of the entropy law, we propose a quite efficient and universal data selection method named \textbf{ZIP} for training LLMs, which aim to prioritize data subsets exhibiting a low compression ratio. Based on a multi-stage algorithm that selects diverse data in a greedy manner, we can obtain a good data subset with satisfactory diversity. Extensive experiments have been conducted to validate the entropy law and the superiority of ZIP across different LLM backbones and alignment stages. We also present an interesting application of entropy law that can detect potential performance risks at the beginning of model training.
Related papers
- Regurgitative Training: The Value of Real Data in Training Large Language Models [1.2815904071470703]
We evaluate the implications of "regurgitative training" on LLM performance.
We find strong evidence that regurgitative training clearly handicaps the performance of LLMs.
We propose and evaluate three different strategies to mitigate the performance loss of regurgitative training.
arXiv Detail & Related papers (2024-07-03T18:42:55Z) - Measuring Sample Importance in Data Pruning for Training LLMs from a Data Compression Perspective [4.079147243688765]
We consider data pruning as a method of data-efficient training of large language models.
We leverage log-likelihood function of trained models as a surrogate to measure information content of samples.
arXiv Detail & Related papers (2024-06-20T09:09:34Z) - LLM2LLM: Boosting LLMs with Novel Iterative Data Enhancement [79.31084387589968]
Pretrained large language models (LLMs) are currently state-of-the-art for solving the vast majority of natural language processing tasks.
We propose LLM2LLM, a data augmentation strategy that uses a teacher LLM to enhance a small seed dataset.
We achieve improvements up to 24.2% on the GSM8K dataset, 32.6% on CaseHOLD, 32.0% on SNIPS, 52.6% on TREC and 39.8% on SST-2 over regular fine-tuning in the low-data regime.
arXiv Detail & Related papers (2024-03-22T08:57:07Z) - 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) - LLMaAA: Making Large Language Models as Active Annotators [32.57011151031332]
We propose LLMaAA, which takes large language models as annotators and puts them into an active learning loop to determine what to annotate efficiently.
We conduct experiments and analysis on two classic NLP tasks, named entity recognition and relation extraction.
With LLMaAA, task-specific models trained from LLM-generated labels can outperform the teacher within only hundreds of annotated examples.
arXiv Detail & Related papers (2023-10-30T14:54:15Z) - Data-Juicer: A One-Stop Data Processing System for Large Language Models [73.27731037450995]
A data recipe is a mixture of data from different sources for training Large Language Models (LLMs)
We build a new system named Data-Juicer, with which we can efficiently generate diverse data recipes.
The data recipes derived with Data-Juicer gain notable improvements on state-of-the-art LLMs.
arXiv Detail & Related papers (2023-09-05T08:22:07Z) - MLLM-DataEngine: An Iterative Refinement Approach for MLLM [62.30753425449056]
We propose a novel closed-loop system that bridges data generation, model training, and evaluation.
Within each loop, the MLLM-DataEngine first analyze the weakness of the model based on the evaluation results.
For targeting, we propose an Adaptive Bad-case Sampling module, which adjusts the ratio of different types of data.
For quality, we resort to GPT-4 to generate high-quality data with each given data type.
arXiv Detail & Related papers (2023-08-25T01:41:04Z) - From Quantity to Quality: Boosting LLM Performance with Self-Guided Data Selection for Instruction Tuning [52.257422715393574]
We introduce a self-guided methodology for Large Language Models (LLMs) to autonomously discern and select cherry samples from open-source datasets.
Our key innovation, the Instruction-Following Difficulty (IFD) metric, emerges as a pivotal metric to identify discrepancies between a model's expected responses and its intrinsic generation capability.
arXiv Detail & Related papers (2023-08-23T09:45:29Z)
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.