超越規模：多樣性係數作為數據質量指標展示了LLM在形式多樣的數據上進行預訓練

摘要

目前預訓練具備能力的大型語言模型（LLMs）的最新趨勢主要集中在模型和數據集大小的擴展上。然而，預訓練數據的質量對於訓練強大的LLMs是一個重要因素，但這是一個尚未完全表徵的模糊概念。因此，我們使用最近提出的Task2Vec多樣性係數來確立和理解數據質量的形式方面，以超越單純的規模。具體來說，我們測量公開可用的預訓練數據集的多樣性係數，以證明它們的形式多樣性與理論下限和上限相比是高的。此外，為了建立對多樣性係數的信心，我們進行解釋性實驗，並發現該係數符合多樣性的直觀特性，例如，隨著潛在概念數量的增加而增加。我們得出結論，多樣性係數是可靠的，顯示其在公開可用的LLM數據集中很高，並推測它可以用於構建對LLMs有用的多樣性數據集。

English

Current trends to pre-train capable Large Language Models (LLMs) mostly focus on scaling of model and dataset size. However, the quality of pre-training data is an important factor for training powerful LLMs, yet it is a nebulous concept that has not been fully characterized. Therefore, we use the recently proposed Task2Vec diversity coefficient to ground and understand formal aspects of data quality, to go beyond scale alone. Specifically, we measure the diversity coefficient of publicly available pre-training datasets to demonstrate that their formal diversity is high when compared to theoretical lower and upper bounds. In addition, to build confidence in the diversity coefficient, we conduct interpretability experiments and find that the coefficient aligns with intuitive properties of diversity, e.g., it increases as the number of latent concepts increases. We conclude the diversity coefficient is reliable, show it's high for publicly available LLM datasets, and conjecture it can be used to build useful diverse datasets for LLMs.

超越規模：多樣性係數作為數據質量指標展示了LLM在形式多樣的數據上進行預訓練

Beyond Scale: the Diversity Coefficient as a Data Quality Metric Demonstrates LLMs are Pre-trained on Formally Diverse Data

摘要

Support