Augmenting In-Context-Learning in LLMs via Automatic Data Labeling and Refinement

TL;DR

This paper introduces ADLR, a method that automatically generates and refines intermediate-step demonstrations to enhance in-context learning in LLMs, significantly improving performance in code-based table QA and mathematical reasoning tasks.

Contribution

The paper presents a novel automatic data labeling and refinement technique to generate intermediate demonstrations, reducing manual effort and improving LLM performance.

Findings

Achieved up to 5.5% performance gain in code-based table QA.

Effective automatic generation and filtering of demonstrations.

Enhanced in-context learning with minimal manual examples.

Abstract

It has been shown that Large Language Models' (LLMs) performance can be improved for many tasks using Chain of Thought (CoT) or In-Context Learning (ICL), which involve demonstrating the steps needed to solve a task using a few examples. However, while datasets with input-output pairs are relatively easy to produce, providing demonstrations which include intermediate steps requires cumbersome manual work. These steps may be executable programs, as in agentic flows, or step-by-step reasoning as in CoT. In this work, we propose Automatic Data Labeling and Refinement (ADLR), a method to automatically generate and filter demonstrations which include the above intermediate steps, starting from a small seed of manually crafted examples. We demonstrate the advantage of ADLR in code-based table QA and mathematical reasoning, achieving up to a 5.5% gain. The code implementing our method is provided in the Supplementary material and will be made available.