Don't Just Demo, Teach Me the Principles: A Principle-Based Multi-Agent Prompting Strategy for Text Classification

2025-02-11

AI Summary

To address the reliance on manually crafted examples, poor generalizability, and high inference cost in few-shot/zero-shot text classification, this paper proposes a principle-induction-based multi-agent prompting framework. Multiple LLM agents collaboratively extract, validate, and aggregate transferable classification principles from minimal or zero labeled data; a dedicated classification agent then performs inference using these principles. Our approach eliminates manual example and principle design, introducing the first closed-loop paradigm of “principle generation → aggregation → application.” Evaluated across diverse domain datasets, it achieves macro-F1 improvements of 1.55–19.37 percentage points over strong baselines—including chain-of-thought and backtrack prompting—while incurring lower inference overhead than standard few-shot prompting and demonstrating superior cross-task generalization.

Technology Category

Multiagent Systems: Modeling other AgentsMachine Learning: Multi-class/Multi-label Learning & Extreme ClassificationNatural Language Processing: Prompt Engineering / Prompting

Application Category

Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactionsSearch and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved informationEconomics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

We present PRINCIPLE-BASED PROMPTING, a simple but effective multi-agent prompting strategy for text classification. It first asks multiple LLM agents to independently generate candidate principles based on analysis of demonstration samples with or without labels, consolidates them into final principles via a finalizer agent, and then sends them to a classifier agent to perform downstream classification tasks. Extensive experiments on binary and multi-class classification datasets with different sizes of LLMs show that our approach not only achieves substantial performance gains (1.55% - 19.37%) over zero-shot prompting on macro-F1 score but also outperforms other strong baselines (CoT and stepback prompting). Principles generated by our approach help LLMs perform better on classification tasks than human crafted principles on two private datasets. Our multi-agent PRINCIPLE-BASED PROMPTING approach also shows on-par or better performance compared to demonstration-based few-shot prompting approaches, yet with substantially lower inference costs. Ablation studies show that label information and the multi-agent cooperative LLM framework play an important role in generating high-quality principles to facilitate downstream classification tasks.

Problem

• Develops multi-agent prompting for text classification.
• Generates principles to enhance LLM classification performance.
• Reduces inference costs compared to few-shot methods.

Innovation

• Multi-agent prompting strategy
• Principle-based text classification
• LLM cooperative framework