Adam's Law: Textual Frequency Law on Large Language Models

TL;DR

This paper introduces a new framework for improving large language models by leveraging textual frequency, including a law, distillation, and curriculum training, validated on diverse reasoning and translation tasks.

Contribution

It proposes the Textual Frequency Law, a novel approach to enhance LLMs through frequency-based prompting, fine-tuning, and data augmentation techniques.

Findings

Frequency-based prompting improves LLM performance.

Curriculum training with increasing frequency enhances model fine-tuning.

The framework shows effectiveness across multiple NLP tasks.

Abstract

While textual frequency has been validated as relevant to human cognition in reading speed, its relatedness to Large Language Models (LLMs) is seldom studied. We propose a novel research direction in terms of textual data frequency, which is an understudied topic, to the best of our knowledge. Our framework is composed of three units. First, this paper proposes Textual Frequency Law (TFL), which indicates that frequent textual data should be preferred for LLMs for both prompting and fine-tuning. Since many LLMs are closed-source in their training data, we propose using online resources to estimate the sentence-level frequency. We then utilize an input paraphraser to paraphrase the input into a more frequent textual expression. Next, we propose Textual Frequency Distillation (TFD) by querying LLMs to conduct story completion by further extending the sentences in the datasets, and the resulting corpora are used to adjust the initial estimation. Finally, we propose Curriculum Textual Frequency Training (CTFT) that fine-tunes LLMs in an increasing order of sentence-level frequency. Experiments are conducted on our curated dataset Textual Frequency Paired Dataset (TFPD) on math reasoning, machine translation, commonsense reasoning and agentic tool calling. Results show the effectiveness of our framework.